diff --git a/apps/x/apps/main/src/main.ts b/apps/x/apps/main/src/main.ts
index 34363b28..b3868bc5 100644
--- a/apps/x/apps/main/src/main.ts
+++ b/apps/x/apps/main/src/main.ts
@@ -17,6 +17,8 @@ import { init as initCalendarSync } from "@x/core/dist/knowledge/sync_calendar.j
 import { init as initFirefliesSync } from "@x/core/dist/knowledge/sync_fireflies.js";
 import { init as initGranolaSync } from "@x/core/dist/knowledge/granola/sync.js";
 import { init as initGraphBuilder } from "@x/core/dist/knowledge/build_graph.js";
+import { init as initEmailLabeling } from "@x/core/dist/knowledge/label_emails.js";
+import { init as initNoteTagging } from "@x/core/dist/knowledge/tag_notes.js";
 import { init as initAgentRunner } from "@x/core/dist/agent-schedule/runner.js";
 import { initConfigs } from "@x/core/dist/config/initConfigs.js";
 import started from "electron-squirrel-startup";
@@ -170,6 +172,12 @@ app.whenReady().then(async () => {
   // start knowledge graph builder
   initGraphBuilder();
 
+  // start email labeling service
+  initEmailLabeling();
+
+  // start note tagging service
+  initNoteTagging();
+
   // start background agent runner (scheduled agents)
   initAgentRunner();
 
diff --git a/apps/x/apps/renderer/src/App.tsx b/apps/x/apps/renderer/src/App.tsx
index 92f43577..fe2f5dd1 100644
--- a/apps/x/apps/renderer/src/App.tsx
+++ b/apps/x/apps/renderer/src/App.tsx
@@ -46,6 +46,7 @@ import {
 import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from "@/components/ui/tooltip"
 import { Toaster } from "@/components/ui/sonner"
 import { stripKnowledgePrefix, toKnowledgePath, wikiLabel } from '@/lib/wiki-links'
+import { splitFrontmatter, joinFrontmatter, extractTags } from '@/lib/frontmatter'
 import { OnboardingModal } from '@/components/onboarding-modal'
 import { SearchDialog } from '@/components/search-dialog'
 import { BackgroundTaskDetail } from '@/components/background-task-detail'
@@ -509,6 +510,10 @@ function App() {
   const initialContentRef = useRef<string>('')
   const renameInProgressRef = useRef(false)
 
+  // Frontmatter state: store raw frontmatter per file path, tags for active file
+  const frontmatterByPathRef = useRef<Map<string, string | null>>(new Map())
+  const [activeFileTags, setActiveFileTags] = useState<string[]>([])
+
   // Version history state
   const [versionHistoryPath, setVersionHistoryPath] = useState<string | null>(null)
   const [viewingHistoricalVersion, setViewingHistoricalVersion] = useState<{
@@ -892,12 +897,15 @@ function App() {
           const result = await window.ipc.invoke('workspace:readFile', { path: pathToReload })
           if (selectedPathRef.current !== pathToReload) return
           setFileContent(result.data)
-          setEditorContent(result.data)
-          setEditorCacheForPath(pathToReload, result.data)
-          editorContentRef.current = result.data
+          const { raw: fm, body } = splitFrontmatter(result.data)
+          frontmatterByPathRef.current.set(pathToReload, fm)
+          setEditorContent(body)
+          setEditorCacheForPath(pathToReload, body)
+          editorContentRef.current = body
           editorPathRef.current = pathToReload
-          initialContentByPathRef.current.set(pathToReload, result.data)
-          initialContentRef.current = result.data
+          initialContentByPathRef.current.set(pathToReload, body)
+          initialContentRef.current = body
+          setActiveFileTags(extractTags(fm))
         }
       }
     })
@@ -926,6 +934,7 @@ function App() {
         editorContentRef.current = cachedContent
         editorPathRef.current = selectedPath
         initialContentRef.current = initialContentByPathRef.current.get(selectedPath) ?? cachedContent
+        setActiveFileTags(extractTags(frontmatterByPathRef.current.get(selectedPath) ?? null))
         return
       }
     }
@@ -940,6 +949,8 @@ function App() {
           const result = await window.ipc.invoke('workspace:readFile', { path: pathToLoad })
           if (cancelled || fileLoadRequestIdRef.current !== requestId || selectedPathRef.current !== pathToLoad) return
           setFileContent(result.data)
+          const { raw: fm, body } = splitFrontmatter(result.data)
+          frontmatterByPathRef.current.set(pathToLoad, fm)
           const normalizeForCompare = (s: string) => s.split('\n').map(line => line.trimEnd()).join('\n').trim()
           const isSameEditorFile = editorPathRef.current === pathToLoad
           const knownBaseline = initialContentByPathRef.current.get(pathToLoad)
@@ -949,15 +960,16 @@ function App() {
             && normalizeForCompare(editorContentRef.current) !== normalizeForCompare(knownBaseline)
           const shouldPreserveActiveDraft = isSameEditorFile && hasUnsavedEdits
           if (!shouldPreserveActiveDraft) {
-            setEditorContent(result.data)
+            setEditorContent(body)
             if (pathToLoad.endsWith('.md')) {
-              setEditorCacheForPath(pathToLoad, result.data)
+              setEditorCacheForPath(pathToLoad, body)
             }
-            editorContentRef.current = result.data
+            editorContentRef.current = body
             editorPathRef.current = pathToLoad
-            initialContentByPathRef.current.set(pathToLoad, result.data)
-            initialContentRef.current = result.data
+            initialContentByPathRef.current.set(pathToLoad, body)
+            initialContentRef.current = body
             setLastSaved(null)
+            setActiveFileTags(extractTags(fm))
           } else {
             // Still update the editor's path so subsequent autosaves write to the correct file.
             editorPathRef.current = pathToLoad
@@ -1006,7 +1018,7 @@ function App() {
       const wasActiveAtStart = selectedPathRef.current === pathAtStart
       if (wasActiveAtStart) setIsSaving(true)
       let pathToSave = pathAtStart
-      let contentToSave = debouncedContent
+      let contentToSave = joinFrontmatter(frontmatterByPathRef.current.get(pathAtStart) ?? null, debouncedContent)
       let renamedFrom: string | null = null
       let renamedTo: string | null = null
       try {
@@ -1036,16 +1048,21 @@ function App() {
                 renameInProgressRef.current = true
                 await window.ipc.invoke('workspace:rename', { from: pathAtStart, to: targetPath })
                 pathToSave = targetPath
-                contentToSave = rewriteWikiLinksForRenamedFileInMarkdown(
+                const rewrittenBody = rewriteWikiLinksForRenamedFileInMarkdown(
                   debouncedContent,
                   pathAtStart,
                   targetPath
                 )
+                contentToSave = joinFrontmatter(frontmatterByPathRef.current.get(pathAtStart) ?? null, rewrittenBody)
                 renamedFrom = pathAtStart
                 renamedTo = targetPath
                 editorPathRef.current = targetPath
                 untitledRenameReadyPathsRef.current.delete(pathAtStart)
                 setFileTabs(prev => prev.map(tab => (tab.path === pathAtStart ? { ...tab, path: targetPath } : tab)))
+                // Migrate frontmatter entry
+                const fmEntry = frontmatterByPathRef.current.get(pathAtStart)
+                frontmatterByPathRef.current.delete(pathAtStart)
+                frontmatterByPathRef.current.set(targetPath, fmEntry ?? null)
                 initialContentByPathRef.current.delete(pathAtStart)
                 const cachedContent = editorContentByPathRef.current.get(pathAtStart)
                 if (cachedContent !== undefined) {
@@ -1070,8 +1087,9 @@ function App() {
                   })
                 }
                 if (selectedPathRef.current === pathAtStart) {
-                  editorContentRef.current = contentToSave
-                  setEditorContent(contentToSave)
+                  const bodyForEditor = splitFrontmatter(contentToSave).body
+                  editorContentRef.current = bodyForEditor
+                  setEditorContent(bodyForEditor)
                 }
               }
             }
@@ -1083,7 +1101,8 @@ function App() {
           opts: { encoding: 'utf8' }
         })
         markRecentLocalMarkdownWrite(pathToSave)
-        initialContentByPathRef.current.set(pathToSave, contentToSave)
+        // Store body-only baseline (matches what debouncedContent compares against)
+        initialContentByPathRef.current.set(pathToSave, splitFrontmatter(contentToSave).body)
 
         // If we renamed the active file, update state/history AFTER the write completes so the editor
         // doesn't reload stale on-disk content mid-typing (which can drop the latest character).
@@ -1104,7 +1123,7 @@ function App() {
 
         // Only update "current file" UI state if we're still on this file
         if (selectedPathRef.current === pathAtStart || selectedPathRef.current === pathToSave) {
-          initialContentRef.current = contentToSave
+          initialContentRef.current = splitFrontmatter(contentToSave).body
           setLastSaved(new Date())
         }
       } catch (err) {
@@ -2162,6 +2181,7 @@ function App() {
       removeEditorCacheForPath(closingTab.path)
       initialContentByPathRef.current.delete(closingTab.path)
       untitledRenameReadyPathsRef.current.delete(closingTab.path)
+      frontmatterByPathRef.current.delete(closingTab.path)
       if (editorPathRef.current === closingTab.path) {
         editorPathRef.current = null
       }
@@ -2768,6 +2788,12 @@ function App() {
         if (editorPathRef.current === oldPath) {
           editorPathRef.current = newPath
         }
+        // Migrate frontmatter entry
+        const fmEntry = frontmatterByPathRef.current.get(oldPath)
+        if (fmEntry !== undefined) {
+          frontmatterByPathRef.current.delete(oldPath)
+          frontmatterByPathRef.current.set(newPath, fmEntry)
+        }
         const baseline = initialContentByPathRef.current.get(oldPath)
         if (baseline !== undefined) {
           initialContentByPathRef.current.delete(oldPath)
@@ -2805,6 +2831,7 @@ function App() {
           removeEditorCacheForPath(path)
           initialContentByPathRef.current.delete(path)
           untitledRenameReadyPathsRef.current.delete(path)
+          frontmatterByPathRef.current.delete(path)
         }
         // Close any file tab showing the deleted file
         const tabForFile = fileTabs.find(t => t.path === path)
@@ -3391,6 +3418,7 @@ function App() {
                               wikiLinks={wikiLinkConfig}
                               onImageUpload={handleImageUpload}
                               editorSessionKey={editorSessionByTabId[tab.id] ?? 0}
+                              tags={isActive ? activeFileTags : undefined}
                               onHistoryHandlersChange={(handlers) => {
                                 if (handlers) {
                                   fileHistoryHandlersRef.current.set(tab.id, handlers)
diff --git a/apps/x/apps/renderer/src/components/bases-view.tsx b/apps/x/apps/renderer/src/components/bases-view.tsx
new file mode 100644
index 00000000..7403435e
--- /dev/null
+++ b/apps/x/apps/renderer/src/components/bases-view.tsx
@@ -0,0 +1,310 @@
+import * as React from 'react'
+import { useEffect, useState, useMemo, useCallback, useRef } from 'react'
+import { ArrowDown, ArrowUp, X } from 'lucide-react'
+import { Badge } from '@/components/ui/badge'
+import { cn } from '@/lib/utils'
+import { splitFrontmatter, extractTags } from '@/lib/frontmatter'
+
+interface TreeNode {
+  path: string
+  name: string
+  kind: 'file' | 'dir'
+  children?: TreeNode[]
+  stat?: { size: number; mtimeMs: number }
+}
+
+type NoteEntry = {
+  path: string
+  name: string
+  folder: string
+  tags: string[]
+  mtimeMs: number
+}
+
+type SortField = 'name' | 'folder' | 'mtimeMs'
+type SortDir = 'asc' | 'desc'
+
+type BasesViewProps = {
+  tree: TreeNode[]
+  onSelectNote: (path: string) => void
+}
+
+function collectFilePaths(nodes: TreeNode[]): { path: string; name: string; mtimeMs: number }[] {
+  return nodes.flatMap((n) =>
+    n.kind === 'file' && n.name.endsWith('.md')
+      ? [{ path: n.path, name: n.name.replace(/\.md$/i, ''), mtimeMs: n.stat?.mtimeMs ?? 0 }]
+      : n.children
+        ? collectFilePaths(n.children)
+        : [],
+  )
+}
+
+/** Build a stable fingerprint from the tree's file paths + mtimes so we only reload when files actually change. */
+function treeFingerprint(nodes: TreeNode[]): string {
+  const files = collectFilePaths(nodes)
+  return files.map((f) => `${f.path}:${f.mtimeMs}`).join('\n')
+}
+
+function getFolder(path: string): string {
+  const parts = path.split('/')
+  if (parts.length >= 3) return parts[1]
+  return ''
+}
+
+function formatDate(ms: number): string {
+  if (!ms) return ''
+  const d = new Date(ms)
+  return d.toLocaleDateString(undefined, { month: 'short', day: 'numeric', year: 'numeric' })
+}
+
+export function BasesView({ tree, onSelectNote }: BasesViewProps) {
+  const [notes, setNotes] = useState<NoteEntry[]>([])
+  const [initialLoading, setInitialLoading] = useState(true)
+  const [selectedTags, setSelectedTags] = useState<Set<string>>(new Set())
+  const [sortField, setSortField] = useState<SortField>('mtimeMs')
+  const [sortDir, setSortDir] = useState<SortDir>('desc')
+  const lastFingerprintRef = useRef<string>('')
+
+  // Stable fingerprint — only changes when actual file paths/mtimes differ
+  const fingerprint = useMemo(() => treeFingerprint(tree), [tree])
+
+  // Load notes data when fingerprint changes
+  useEffect(() => {
+    if (fingerprint === lastFingerprintRef.current) return
+    lastFingerprintRef.current = fingerprint
+
+    let cancelled = false
+    const files = collectFilePaths(tree)
+
+    async function loadNotes() {
+      const entries: NoteEntry[] = []
+
+      for (const file of files) {
+        try {
+          const result = await window.ipc.invoke('workspace:readFile', {
+            path: file.path,
+            encoding: 'utf8',
+          })
+          const { raw } = splitFrontmatter(result.data)
+          const tags = extractTags(raw)
+          entries.push({
+            path: file.path,
+            name: file.name,
+            folder: getFolder(file.path),
+            tags,
+            mtimeMs: file.mtimeMs,
+          })
+        } catch {
+          entries.push({
+            path: file.path,
+            name: file.name,
+            folder: getFolder(file.path),
+            tags: [],
+            mtimeMs: file.mtimeMs,
+          })
+        }
+      }
+
+      if (!cancelled) {
+        setNotes(entries)
+        setInitialLoading(false)
+      }
+    }
+
+    loadNotes()
+    return () => { cancelled = true }
+  }, [fingerprint, tree])
+
+  // Collect all unique tags
+  const allTags = useMemo(() => {
+    const tagSet = new Set<string>()
+    for (const note of notes) {
+      for (const tag of note.tags) {
+        tagSet.add(tag)
+      }
+    }
+    return [...tagSet].sort((a, b) => a.localeCompare(b))
+  }, [notes])
+
+  // Filter and sort
+  const filteredNotes = useMemo(() => {
+    let result = notes
+    if (selectedTags.size > 0) {
+      const tagsArray = [...selectedTags]
+      result = result.filter((note) =>
+        tagsArray.every((tag) => note.tags.includes(tag)),
+      )
+    }
+    result = [...result].sort((a, b) => {
+      let cmp = 0
+      if (sortField === 'name') {
+        cmp = a.name.localeCompare(b.name)
+      } else if (sortField === 'folder') {
+        cmp = a.folder.localeCompare(b.folder)
+      } else {
+        cmp = a.mtimeMs - b.mtimeMs
+      }
+      return sortDir === 'asc' ? cmp : -cmp
+    })
+    return result
+  }, [notes, selectedTags, sortField, sortDir])
+
+  const toggleTag = useCallback((tag: string) => {
+    setSelectedTags((prev) => {
+      const next = new Set(prev)
+      if (next.has(tag)) {
+        next.delete(tag)
+      } else {
+        next.add(tag)
+      }
+      return next
+    })
+  }, [])
+
+  const clearFilters = useCallback(() => {
+    setSelectedTags(new Set())
+  }, [])
+
+  const handleSort = useCallback((field: SortField) => {
+    setSortField((prev) => {
+      if (prev === field) {
+        setSortDir((d) => (d === 'asc' ? 'desc' : 'asc'))
+        return prev
+      }
+      setSortDir(field === 'mtimeMs' ? 'desc' : 'asc')
+      return field
+    })
+  }, [])
+
+  const SortIcon = ({ field }: { field: SortField }) => {
+    if (sortField !== field) return null
+    return sortDir === 'asc' ? (
+      <ArrowUp className="size-3 inline ml-1" />
+    ) : (
+      <ArrowDown className="size-3 inline ml-1" />
+    )
+  }
+
+  if (initialLoading) {
+    return (
+      <div className="flex-1 flex items-center justify-center">
+        <div className="space-y-3 w-full max-w-2xl px-6">
+          {Array.from({ length: 8 }).map((_, i) => (
+            <div key={i} className="h-10 rounded bg-muted/50 animate-pulse" />
+          ))}
+        </div>
+      </div>
+    )
+  }
+
+  return (
+    <div className="flex-1 flex flex-col overflow-hidden">
+      {/* Filter bar */}
+      <div className="shrink-0 border-b border-border px-4 py-2.5">
+        <div className="flex items-center gap-2 flex-wrap">
+          <span className="text-xs text-muted-foreground shrink-0">
+            Showing {filteredNotes.length} of {notes.length} notes
+          </span>
+          {selectedTags.size > 0 && (
+            <button
+              onClick={clearFilters}
+              className="text-xs text-muted-foreground hover:text-foreground flex items-center gap-1 shrink-0"
+            >
+              <X className="size-3" />
+              Clear filters
+            </button>
+          )}
+          <div className="flex items-center gap-1 flex-wrap">
+            {allTags.map((tag) => (
+              <button
+                key={tag}
+                onClick={() => toggleTag(tag)}
+                className={cn(
+                  'inline-flex items-center rounded-full border px-2 py-0.5 text-[11px] font-medium transition-colors',
+                  selectedTags.has(tag)
+                    ? 'bg-primary text-primary-foreground border-transparent'
+                    : 'bg-transparent text-muted-foreground border-border hover:bg-accent hover:text-accent-foreground',
+                )}
+              >
+                {tag}
+              </button>
+            ))}
+          </div>
+        </div>
+      </div>
+
+      {/* Table */}
+      <div className="flex-1 overflow-auto">
+        <table className="w-full text-sm">
+          <thead className="sticky top-0 bg-background border-b border-border">
+            <tr>
+              <th
+                className="text-left px-4 py-2 font-medium text-muted-foreground cursor-pointer hover:text-foreground select-none"
+                onClick={() => handleSort('name')}
+              >
+                Name
+                <SortIcon field="name" />
+              </th>
+              <th
+                className="text-left px-4 py-2 font-medium text-muted-foreground cursor-pointer hover:text-foreground select-none"
+                onClick={() => handleSort('folder')}
+              >
+                Folder
+                <SortIcon field="folder" />
+              </th>
+              <th className="text-left px-4 py-2 font-medium text-muted-foreground select-none">
+                Tags
+              </th>
+              <th
+                className="text-left px-4 py-2 font-medium text-muted-foreground cursor-pointer hover:text-foreground select-none"
+                onClick={() => handleSort('mtimeMs')}
+              >
+                Last Modified
+                <SortIcon field="mtimeMs" />
+              </th>
+            </tr>
+          </thead>
+          <tbody>
+            {filteredNotes.map((note) => (
+              <tr
+                key={note.path}
+                className="border-b border-border/50 hover:bg-accent/50 cursor-pointer transition-colors"
+                onClick={() => onSelectNote(note.path)}
+              >
+                <td className="px-4 py-2 font-medium">{note.name}</td>
+                <td className="px-4 py-2 text-muted-foreground">{note.folder}</td>
+                <td className="px-4 py-2">
+                  <div className="flex items-center gap-1 flex-wrap">
+                    {note.tags.map((tag) => (
+                      <Badge
+                        key={tag}
+                        variant="secondary"
+                        className="text-[10px] px-1.5 py-0 cursor-pointer hover:bg-primary hover:text-primary-foreground"
+                        onClick={(e) => {
+                          e.stopPropagation()
+                          toggleTag(tag)
+                        }}
+                      >
+                        {tag}
+                      </Badge>
+                    ))}
+                  </div>
+                </td>
+                <td className="px-4 py-2 text-muted-foreground whitespace-nowrap">
+                  {formatDate(note.mtimeMs)}
+                </td>
+              </tr>
+            ))}
+            {filteredNotes.length === 0 && (
+              <tr>
+                <td colSpan={4} className="px-4 py-8 text-center text-muted-foreground">
+                  No notes found
+                </td>
+              </tr>
+            )}
+          </tbody>
+        </table>
+      </div>
+    </div>
+  )
+}
diff --git a/apps/x/apps/renderer/src/components/markdown-editor.tsx b/apps/x/apps/renderer/src/components/markdown-editor.tsx
index 37a827cb..35cab547 100644
--- a/apps/x/apps/renderer/src/components/markdown-editor.tsx
+++ b/apps/x/apps/renderer/src/components/markdown-editor.tsx
@@ -176,6 +176,7 @@ function getMarkdownWithBlankLines(editor: Editor): string {
   return result
 }
 import { EditorToolbar } from './editor-toolbar'
+import { TagPills } from './tag-pills'
 import { WikiLink } from '@/extensions/wiki-link'
 import { Popover, PopoverAnchor, PopoverContent } from '@/components/ui/popover'
 import { Command, CommandEmpty, CommandItem, CommandList } from '@/components/ui/command'
@@ -200,6 +201,7 @@ interface MarkdownEditorProps {
   editorSessionKey?: number
   onHistoryHandlersChange?: (handlers: { undo: () => boolean; redo: () => boolean } | null) => void
   editable?: boolean
+  tags?: string[]
 }
 
 type WikiLinkMatch = {
@@ -288,6 +290,7 @@ export function MarkdownEditor({
   editorSessionKey = 0,
   onHistoryHandlersChange,
   editable = true,
+  tags,
 }: MarkdownEditorProps) {
   const isInternalUpdate = useRef(false)
   const wrapperRef = useRef<HTMLDivElement>(null)
@@ -721,6 +724,7 @@ export function MarkdownEditor({
         onSelectionHighlight={setSelectionHighlight}
         onImageUpload={handleImageUploadWithPlaceholder}
       />
+      {tags && <TagPills tags={tags} />}
       <div className="editor-content-wrapper" ref={wrapperRef} onScroll={handleScroll}>
         <EditorContent editor={editor} />
         {wikiLinks ? (
diff --git a/apps/x/apps/renderer/src/components/settings-dialog.tsx b/apps/x/apps/renderer/src/components/settings-dialog.tsx
index 7cb0a011..fb28e2b0 100644
--- a/apps/x/apps/renderer/src/components/settings-dialog.tsx
+++ b/apps/x/apps/renderer/src/components/settings-dialog.tsx
@@ -1,8 +1,8 @@
 "use client"
 
 import * as React from "react"
-import { useState, useEffect, useCallback } from "react"
-import { Server, Key, Shield, Palette, Monitor, Sun, Moon, Loader2, CheckCircle2, Plus, X } from "lucide-react"
+import { useState, useEffect, useCallback, useMemo } from "react"
+import { Server, Key, Shield, Palette, Monitor, Sun, Moon, Loader2, CheckCircle2, Tags, ChevronRight, Plus, X } from "lucide-react"
 
 import {
   Dialog,
@@ -22,7 +22,7 @@ import { cn } from "@/lib/utils"
 import { useTheme } from "@/contexts/theme-context"
 import { toast } from "sonner"
 
-type ConfigTab = "models" | "mcp" | "security" | "appearance"
+type ConfigTab = "models" | "mcp" | "security" | "appearance" | "note-tagging"
 
 interface TabConfig {
   id: ConfigTab
@@ -60,6 +60,13 @@ const tabs: TabConfig[] = [
     icon: Palette,
     description: "Customize the look and feel",
   },
+  {
+    id: "note-tagging",
+    label: "Note Tagging",
+    icon: Tags,
+    path: "config/tags.json",
+    description: "Configure tags for notes and emails",
+  },
 ]
 
 interface SettingsDialogProps {
@@ -685,6 +692,311 @@ function ModelSettings({ dialogOpen }: { dialogOpen: boolean }) {
   )
 }
 
+// --- Note Tagging Settings ---
+
+interface TagDef {
+  tag: string
+  type: string
+  applicability: "email" | "notes" | "both"
+  description: string
+  example?: string
+}
+
+const TAG_TYPE_ORDER = [
+  "relationship", "relationship-sub", "topic", "email-type",
+  "filter", "action", "status", "source",
+]
+
+const TAG_TYPE_LABELS: Record<string, string> = {
+  "relationship": "Relationship",
+  "relationship-sub": "Relationship Sub-Tags",
+  "topic": "Topic",
+  "email-type": "Email Type",
+  "filter": "Filter",
+  "action": "Action",
+  "status": "Status",
+  "source": "Source",
+}
+
+const DEFAULT_TAGS: TagDef[] = [
+  { tag: "investor", type: "relationship", applicability: "both", description: "Investors, VCs, or angels", example: "Following up on our meeting — we'd like to move forward with the Series A term sheet." },
+  { tag: "customer", type: "relationship", applicability: "both", description: "Paying customers", example: "We're seeing great results with Rowboat. Can we discuss expanding to more teams?" },
+  { tag: "prospect", type: "relationship", applicability: "both", description: "Potential customers", example: "Thanks for the demo yesterday. We're interested in starting a pilot." },
+  { tag: "partner", type: "relationship", applicability: "both", description: "Business partners", example: "Let's discuss how we can promote the integration to both our user bases." },
+  { tag: "vendor", type: "relationship", applicability: "both", description: "Service providers you work with", example: "Here are the updated employment agreements you requested." },
+  { tag: "product", type: "relationship", applicability: "both", description: "Products or services you use (automated)", example: "Your AWS bill for January 2025 is now available." },
+  { tag: "candidate", type: "relationship", applicability: "both", description: "Job applicants", example: "Thanks for reaching out. I'd love to learn more about the engineering role." },
+  { tag: "team", type: "relationship", applicability: "both", description: "Internal team members", example: "Here's the updated roadmap for Q2. Let's discuss in our sync." },
+  { tag: "advisor", type: "relationship", applicability: "both", description: "Advisors, mentors, or board members", example: "I've reviewed the deck. Here are my thoughts on the GTM strategy." },
+  { tag: "personal", type: "relationship", applicability: "both", description: "Family or friends", example: "Are you coming to Thanksgiving this year? Let me know your travel dates." },
+  { tag: "press", type: "relationship", applicability: "both", description: "Journalists or media", example: "I'm writing a piece on AI agents. Would you be available for an interview?" },
+  { tag: "community", type: "relationship", applicability: "both", description: "Users, peers, or open source contributors", example: "Love what you're building with Rowboat. Here's a bug I found..." },
+  { tag: "government", type: "relationship", applicability: "both", description: "Government agencies", example: "Your Delaware franchise tax is due by March 1, 2025." },
+  { tag: "primary", type: "relationship-sub", applicability: "notes", description: "Main contact or decision maker", example: "Sarah Chen — VP Engineering, your main point of contact at Acme." },
+  { tag: "secondary", type: "relationship-sub", applicability: "notes", description: "Supporting contact, involved but not the lead", example: "David Kim — Engineer CC'd on customer emails." },
+  { tag: "executive-assistant", type: "relationship-sub", applicability: "notes", description: "EA or admin handling scheduling and logistics", example: "Lisa — Sarah's EA who schedules all her meetings." },
+  { tag: "cc", type: "relationship-sub", applicability: "notes", description: "Person who's CC'd but not actively engaged", example: "Manager looped in for visibility on deal." },
+  { tag: "referred-by", type: "relationship-sub", applicability: "notes", description: "Person who made an introduction or referral", example: "David Park — Investor who intro'd you to Sarah." },
+  { tag: "former", type: "relationship-sub", applicability: "notes", description: "Previously held this relationship, no longer active", example: "John — Former customer who churned last year." },
+  { tag: "champion", type: "relationship-sub", applicability: "notes", description: "Internal advocate pushing for you", example: "Engineer who loves your product and is selling internally." },
+  { tag: "blocker", type: "relationship-sub", applicability: "notes", description: "Person opposing or blocking progress", example: "CFO resistant to spending on new tools." },
+  { tag: "sales", type: "topic", applicability: "both", description: "Sales conversations, deals, and revenue", example: "Here's the pricing proposal we discussed. Let me know if you have questions." },
+  { tag: "support", type: "topic", applicability: "both", description: "Help requests, issues, and customer support", example: "We're seeing an error when trying to export. Can you help?" },
+  { tag: "legal", type: "topic", applicability: "both", description: "Contracts, terms, compliance, and legal matters", example: "Legal has reviewed the MSA. Attached are our requested changes." },
+  { tag: "finance", type: "topic", applicability: "both", description: "Money, invoices, payments, banking, and taxes", example: "Your invoice #1234 for $5,000 is attached. Payment due in 30 days." },
+  { tag: "hiring", type: "topic", applicability: "both", description: "Recruiting, interviews, and employment", example: "We'd like to move forward with a final round interview. Are you available Thursday?" },
+  { tag: "fundraising", type: "topic", applicability: "both", description: "Raising money and investor relations", example: "Thanks for sending the deck. We'd like to schedule a partner meeting." },
+  { tag: "travel", type: "topic", applicability: "both", description: "Flights, hotels, trips, and travel logistics", example: "Your flight to Tokyo on March 15 is confirmed. Confirmation #ABC123." },
+  { tag: "event", type: "topic", applicability: "both", description: "Conferences, meetups, and gatherings", example: "You're invited to speak at TechCrunch Disrupt. Can you confirm your availability?" },
+  { tag: "shopping", type: "topic", applicability: "both", description: "Purchases, orders, and returns", example: "Your order #12345 has shipped. Track it here." },
+  { tag: "health", type: "topic", applicability: "both", description: "Medical, wellness, and health-related matters", example: "Your appointment with Dr. Smith is confirmed for Monday at 2pm." },
+  { tag: "learning", type: "topic", applicability: "both", description: "Courses, education, and skill-building", example: "Welcome to the Advanced Python course. Here's your access link." },
+  { tag: "research", type: "topic", applicability: "both", description: "Research requests and information gathering", example: "Here's the market analysis you requested on the AI agent space." },
+  { tag: "intro", type: "email-type", applicability: "both", description: "Warm introduction from someone you know", example: "I'd like to introduce you to Sarah Chen, VP Engineering at Acme." },
+  { tag: "followup", type: "email-type", applicability: "both", description: "Following up on a previous conversation", example: "Following up on our call last week. Have you had a chance to review the proposal?" },
+  { tag: "scheduling", type: "email-type", applicability: "email", description: "Meeting and calendar scheduling", example: "Are you available for a call next Tuesday at 2pm?" },
+  { tag: "cold-outreach", type: "email-type", applicability: "email", description: "Unsolicited contact from someone you don't know", example: "Hi, I noticed your company is growing fast. I'd love to show you how we can help with..." },
+  { tag: "newsletter", type: "email-type", applicability: "email", description: "Newsletters, marketing emails, and subscriptions", example: "This week in AI: The latest developments in agent frameworks..." },
+  { tag: "notification", type: "email-type", applicability: "email", description: "Automated alerts, receipts, and system notifications", example: "Your password was changed successfully. If this wasn't you, contact support." },
+  { tag: "spam", type: "filter", applicability: "email", description: "Junk and unwanted email", example: "Congratulations! You've won $1,000,000..." },
+  { tag: "promotion", type: "filter", applicability: "email", description: "Marketing offers and sales pitches", example: "50% off all items this weekend only!" },
+  { tag: "social", type: "filter", applicability: "email", description: "Social media notifications", example: "John Smith commented on your post." },
+  { tag: "forums", type: "filter", applicability: "email", description: "Mailing lists and group discussions", example: "Re: [dev-list] Question about API design" },
+  { tag: "action-required", type: "action", applicability: "both", description: "Needs a response or action from you", example: "Can you send me the pricing by Friday?" },
+  { tag: "fyi", type: "action", applicability: "email", description: "Informational only, no action needed", example: "Just wanted to let you know the deal closed. Thanks for your help!" },
+  { tag: "urgent", type: "action", applicability: "both", description: "Time-sensitive, needs immediate attention", example: "We need your signature on the contract by EOD today or we lose the deal." },
+  { tag: "waiting", type: "action", applicability: "both", description: "Waiting on a response from them" },
+  { tag: "unread", type: "status", applicability: "email", description: "Not yet processed" },
+  { tag: "to-reply", type: "status", applicability: "email", description: "Need to respond" },
+  { tag: "done", type: "status", applicability: "email", description: "Handled, can be archived" },
+  { tag: "active", type: "status", applicability: "notes", description: "Currently relevant, recent activity" },
+  { tag: "archived", type: "status", applicability: "notes", description: "No longer active, kept for reference" },
+  { tag: "stale", type: "status", applicability: "notes", description: "No activity in 60+ days, needs attention or archive" },
+  { tag: "email", type: "source", applicability: "notes", description: "Created or updated from email" },
+  { tag: "meeting", type: "source", applicability: "notes", description: "Created or updated from meeting transcript" },
+  { tag: "browser", type: "source", applicability: "notes", description: "Content captured from web browsing" },
+  { tag: "web-search", type: "source", applicability: "notes", description: "Information from web search" },
+  { tag: "manual", type: "source", applicability: "notes", description: "Manually entered by user" },
+  { tag: "import", type: "source", applicability: "notes", description: "Imported from another system" },
+]
+
+function NoteTaggingSettings({ dialogOpen }: { dialogOpen: boolean }) {
+  const [tags, setTags] = useState<TagDef[]>([])
+  const [originalTags, setOriginalTags] = useState<TagDef[]>([])
+  const [loading, setLoading] = useState(true)
+  const [saving, setSaving] = useState(false)
+  const [collapsedGroups, setCollapsedGroups] = useState<Set<string>>(new Set())
+
+  const hasChanges = JSON.stringify(tags) !== JSON.stringify(originalTags)
+
+  useEffect(() => {
+    if (!dialogOpen) return
+    async function load() {
+      setLoading(true)
+      try {
+        const result = await window.ipc.invoke("workspace:readFile", { path: "config/tags.json" })
+        const parsed = JSON.parse(result.data)
+        setTags(parsed)
+        setOriginalTags(parsed)
+      } catch {
+        setTags([...DEFAULT_TAGS])
+        setOriginalTags([...DEFAULT_TAGS])
+      } finally {
+        setLoading(false)
+      }
+    }
+    load()
+  }, [dialogOpen])
+
+  const groups = useMemo(() => {
+    const map = new Map<string, TagDef[]>()
+    for (const tag of tags) {
+      const list = map.get(tag.type) ?? []
+      list.push(tag)
+      map.set(tag.type, list)
+    }
+    return TAG_TYPE_ORDER.map(type => ({
+      type,
+      label: TAG_TYPE_LABELS[type],
+      tags: map.get(type) ?? [],
+    }))
+  }, [tags])
+
+  const getGlobalIndex = useCallback((type: string, localIndex: number) => {
+    let count = 0
+    for (let i = 0; i < tags.length; i++) {
+      if (tags[i].type === type) {
+        if (count === localIndex) return i
+        count++
+      }
+    }
+    return -1
+  }, [tags])
+
+  const updateTag = useCallback((index: number, field: keyof TagDef, value: string) => {
+    setTags(prev => prev.map((t, i) => i === index ? { ...t, [field]: value } : t))
+  }, [])
+
+  const removeTag = useCallback((index: number) => {
+    setTags(prev => prev.filter((_, i) => i !== index))
+  }, [])
+
+  const addTag = useCallback((type: string) => {
+    const newTag: TagDef = { tag: "", type, applicability: "both", description: "" }
+    const lastIndex = tags.reduce((acc, t, i) => t.type === type ? i : acc, -1)
+    if (lastIndex === -1) {
+      setTags(prev => [...prev, newTag])
+    } else {
+      setTags(prev => [...prev.slice(0, lastIndex + 1), newTag, ...prev.slice(lastIndex + 1)])
+    }
+  }, [tags])
+
+  const handleSave = useCallback(async () => {
+    setSaving(true)
+    try {
+      await window.ipc.invoke("workspace:writeFile", {
+        path: "config/tags.json",
+        data: JSON.stringify(tags, null, 2),
+      })
+      setOriginalTags([...tags])
+      toast.success("Tag configuration saved")
+    } catch {
+      toast.error("Failed to save tag configuration")
+    } finally {
+      setSaving(false)
+    }
+  }, [tags])
+
+  const handleReset = useCallback(() => {
+    if (!confirm("Reset all tags to defaults? This will discard your changes.")) return
+    setTags([...DEFAULT_TAGS])
+  }, [])
+
+  const toggleGroup = useCallback((type: string) => {
+    setCollapsedGroups(prev => {
+      const next = new Set(prev)
+      if (next.has(type)) next.delete(type)
+      else next.add(type)
+      return next
+    })
+  }, [])
+
+  if (loading) {
+    return (
+      <div className="h-full flex items-center justify-center text-muted-foreground text-sm">
+        <Loader2 className="size-4 animate-spin mr-2" />
+        Loading...
+      </div>
+    )
+  }
+
+  return (
+    <div className="h-full flex flex-col">
+      <div className="flex-1 overflow-y-auto space-y-4 min-h-0">
+        {groups.map(group => (
+          <div key={group.type}>
+            <div className="flex items-center justify-between mb-1.5">
+              <button
+                onClick={() => toggleGroup(group.type)}
+                className="flex items-center gap-1 text-xs font-medium uppercase tracking-wider text-muted-foreground hover:text-foreground transition-colors"
+              >
+                <ChevronRight className={cn("size-3.5 transition-transform", !collapsedGroups.has(group.type) && "rotate-90")} />
+                {group.label}
+                <span className="text-[10px] ml-0.5">({group.tags.length})</span>
+              </button>
+              <Button
+                variant="ghost"
+                size="sm"
+                className="h-6 px-2 text-xs"
+                onClick={() => addTag(group.type)}
+              >
+                <Plus className="size-3 mr-1" />
+                Add
+              </Button>
+            </div>
+            {!collapsedGroups.has(group.type) && group.tags.length > 0 && (
+              <div className="border rounded-md overflow-hidden">
+                <div className="grid grid-cols-[100px_1fr_1fr_80px_24px] gap-1 bg-muted/50 px-2 py-1 text-[10px] font-medium text-muted-foreground uppercase tracking-wider">
+                  <div>Tag</div>
+                  <div>Description</div>
+                  <div>Example</div>
+                  <div>Applies to</div>
+                  <div />
+                </div>
+                {group.tags.map((tag, localIdx) => {
+                  const globalIdx = getGlobalIndex(group.type, localIdx)
+                  return (
+                    <div key={globalIdx} className="grid grid-cols-[100px_1fr_1fr_80px_24px] gap-1 border-t px-2 py-0.5 items-center">
+                      <Input
+                        value={tag.tag}
+                        onChange={e => updateTag(globalIdx, "tag", e.target.value)}
+                        className="h-7 text-xs"
+                        placeholder="tag-name"
+                        title={tag.tag}
+                      />
+                      <Input
+                        value={tag.description}
+                        onChange={e => updateTag(globalIdx, "description", e.target.value)}
+                        className="h-7 text-xs"
+                        placeholder="Description"
+                        title={tag.description}
+                      />
+                      <Input
+                        value={tag.example || ""}
+                        onChange={e => updateTag(globalIdx, "example", e.target.value)}
+                        className="h-7 text-xs"
+                        placeholder="Example"
+                        title={tag.example || ""}
+                      />
+                      <Select
+                        value={tag.applicability}
+                        onValueChange={v => updateTag(globalIdx, "applicability", v)}
+                      >
+                        <SelectTrigger className="h-7 text-xs">
+                          <SelectValue />
+                        </SelectTrigger>
+                        <SelectContent>
+                          <SelectItem value="both">Both</SelectItem>
+                          <SelectItem value="email">Email</SelectItem>
+                          <SelectItem value="notes">Notes</SelectItem>
+                        </SelectContent>
+                      </Select>
+                      <button
+                        onClick={() => removeTag(globalIdx)}
+                        className="flex items-center justify-center text-muted-foreground hover:text-destructive transition-colors"
+                      >
+                        <X className="size-3.5" />
+                      </button>
+                    </div>
+                  )
+                })}
+              </div>
+            )}
+            {!collapsedGroups.has(group.type) && group.tags.length === 0 && (
+              <div className="text-xs text-muted-foreground italic px-2">No tags in this group</div>
+            )}
+          </div>
+        ))}
+      </div>
+      <div className="pt-3 border-t mt-3 flex items-center justify-between">
+        <div>
+          {hasChanges && (
+            <span className="text-xs text-muted-foreground">Unsaved changes</span>
+          )}
+        </div>
+        <div className="flex items-center gap-2">
+          <Button variant="outline" size="sm" onClick={handleReset}>
+            Reset to defaults
+          </Button>
+          <Button size="sm" onClick={handleSave} disabled={saving || !hasChanges}>
+            {saving ? "Saving..." : "Save"}
+          </Button>
+        </div>
+      </div>
+    </div>
+  )
+}
+
 // --- Main Settings Dialog ---
 
 export function SettingsDialog({ children }: SettingsDialogProps) {
@@ -708,7 +1020,7 @@ export function SettingsDialog({ children }: SettingsDialogProps) {
   }
 
   const loadConfig = useCallback(async (tab: ConfigTab) => {
-    if (tab === "appearance" || tab === "models") return
+    if (tab === "appearance" || tab === "models" || tab === "note-tagging") return
     const tabConfig = tabs.find((t) => t.id === tab)!
     if (!tabConfig.path) return
     setLoading(true)
@@ -814,9 +1126,11 @@ export function SettingsDialog({ children }: SettingsDialogProps) {
             </div>
 
             {/* Content */}
-            <div className={cn("flex-1 p-4 min-h-0", activeTab === "models" ? "overflow-y-auto" : "overflow-hidden")}>
+            <div className={cn("flex-1 p-4 min-h-0", activeTab === "models" ? "overflow-y-auto" : activeTab === "note-tagging" ? "overflow-hidden flex flex-col" : "overflow-hidden")}>
               {activeTab === "models" ? (
                 <ModelSettings dialogOpen={open} />
+              ) : activeTab === "note-tagging" ? (
+                <NoteTaggingSettings dialogOpen={open} />
               ) : activeTab === "appearance" ? (
                 <AppearanceSettings />
               ) : loading ? (
diff --git a/apps/x/apps/renderer/src/components/tag-pills.tsx b/apps/x/apps/renderer/src/components/tag-pills.tsx
new file mode 100644
index 00000000..eead6558
--- /dev/null
+++ b/apps/x/apps/renderer/src/components/tag-pills.tsx
@@ -0,0 +1,17 @@
+interface TagPillsProps {
+  tags: string[]
+}
+
+export function TagPills({ tags }: TagPillsProps) {
+  if (tags.length === 0) return null
+
+  return (
+    <div className="tag-pills-row">
+      {tags.map((tag, i) => (
+        <span key={`${tag}-${i}`} className="tag-pill">
+          {tag}
+        </span>
+      ))}
+    </div>
+  )
+}
diff --git a/apps/x/apps/renderer/src/lib/frontmatter.ts b/apps/x/apps/renderer/src/lib/frontmatter.ts
new file mode 100644
index 00000000..a9b6b2ff
--- /dev/null
+++ b/apps/x/apps/renderer/src/lib/frontmatter.ts
@@ -0,0 +1,164 @@
+/**
+ * Utilities for splitting, joining, and extracting tags from YAML frontmatter
+ * in knowledge notes and email files.
+ */
+
+/** Split content into raw frontmatter block and body text. */
+export function splitFrontmatter(content: string): { raw: string | null; body: string } {
+  if (!content.startsWith('---')) {
+    return { raw: null, body: content }
+  }
+  const endIndex = content.indexOf('\n---', 3)
+  if (endIndex === -1) {
+    return { raw: null, body: content }
+  }
+  // raw includes both delimiters and the trailing newline after closing ---
+  const closingEnd = endIndex + 4 // '\n---' is 4 chars
+  const raw = content.slice(0, closingEnd)
+  // body starts after the closing --- and its trailing newline
+  let body = content.slice(closingEnd)
+  if (body.startsWith('\n')) {
+    body = body.slice(1)
+  }
+  return { raw, body }
+}
+
+/** Re-prepend raw frontmatter before body when saving. */
+export function joinFrontmatter(raw: string | null, body: string): string {
+  if (!raw) return body
+  return raw + '\n' + body
+}
+
+/** Tag category keys used in the categorized frontmatter format. */
+const TAG_CATEGORY_KEYS = new Set([
+  'relationship',
+  'relationship_sub',
+  'topic',
+  'email_type',
+  'action',
+  'status',
+  'source',
+])
+
+/** Keys that are metadata, not tags — skip when collecting tags. */
+const METADATA_KEYS = new Set(['processed', 'labeled_at', 'tagged_at'])
+
+/**
+ * Extract tags from raw frontmatter YAML.
+ *
+ * Handles three formats:
+ * - Legacy flat list: `tags:` followed by `  - value` items
+ * - Categorized format: top-level keys like `relationship: customer` or
+ *   `topic:` followed by `  - value` list items
+ * - Email format: `labels:` with nested keys (relationship, topics, type, filter, action)
+ *   where values can be single strings or `  - value` arrays
+ *
+ * Skips metadata keys like `processed`, `labeled_at`, `tagged_at`.
+ */
+export function extractTags(raw: string | null): string[] {
+  if (!raw) return []
+
+  const lines = raw.split('\n')
+  const tags: string[] = []
+
+  let inTags = false
+  let inLabels = false
+  let inLabelSubKey = false
+  let inCategoryList = false
+
+  for (const line of lines) {
+    // Top-level key detection — resets all nested state
+    if (/^\w/.test(line) || line === '---') {
+      inTags = false
+      inLabels = false
+      inLabelSubKey = false
+      inCategoryList = false
+    }
+
+    // Legacy note format: tags:
+    if (/^tags:\s*$/.test(line)) {
+      inTags = true
+      inLabels = false
+      inCategoryList = false
+      continue
+    }
+
+    // Email format: labels:
+    if (/^labels:\s*$/.test(line)) {
+      inLabels = true
+      inTags = false
+      inCategoryList = false
+      continue
+    }
+
+    // Categorized format: top-level tag category key
+    const topKeyMatch = line.match(/^(\w+):\s*(.*)$/)
+    if (topKeyMatch) {
+      const key = topKeyMatch[1]
+      const inlineValue = topKeyMatch[2].trim()
+
+      if (TAG_CATEGORY_KEYS.has(key)) {
+        if (inlineValue) {
+          // Single value: `relationship: customer`
+          tags.push(inlineValue)
+          inCategoryList = false
+        } else {
+          // List follows: `topic:\n  - sales`
+          inCategoryList = true
+        }
+        continue
+      }
+    }
+
+    // Collect tag items under `tags:`
+    if (inTags) {
+      const match = line.match(/^\s+-\s+(.+)$/)
+      if (match) {
+        tags.push(match[1].trim())
+      }
+      continue
+    }
+
+    // Collect list items under a category key
+    if (inCategoryList) {
+      const match = line.match(/^\s+-\s+(.+)$/)
+      if (match) {
+        tags.push(match[1].trim())
+      }
+      continue
+    }
+
+    // Handle labels: nested structure
+    if (inLabels) {
+      // Sub-key like `  relationship:` or `  topics:`
+      const subKeyMatch = line.match(/^\s{2}(\w+):\s*(.*)$/)
+      if (subKeyMatch) {
+        const key = subKeyMatch[1]
+        const inlineValue = subKeyMatch[2].trim()
+        if (METADATA_KEYS.has(key)) {
+          inLabelSubKey = false
+          continue
+        }
+        if (inlineValue) {
+          // Inline value like `  type: person`
+          tags.push(inlineValue)
+          inLabelSubKey = false
+        } else {
+          // Array follows
+          inLabelSubKey = true
+        }
+        continue
+      }
+
+      // Array item under a sub-key like `    - value`
+      if (inLabelSubKey) {
+        const itemMatch = line.match(/^\s{4}-\s+(.+)$/)
+        if (itemMatch) {
+          tags.push(itemMatch[1].trim())
+        }
+      }
+    }
+  }
+
+  return tags
+}
diff --git a/apps/x/apps/renderer/src/styles/editor.css b/apps/x/apps/renderer/src/styles/editor.css
index 31ce2bf1..6e1c0deb 100644
--- a/apps/x/apps/renderer/src/styles/editor.css
+++ b/apps/x/apps/renderer/src/styles/editor.css
@@ -237,6 +237,34 @@
   flex-shrink: 0;
 }
 
+/* Tag pills row shown between toolbar and editor content */
+.tag-pills-row {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 4px;
+  padding: 4px 12px;
+  border-bottom: 1px solid var(--border);
+  background-color: var(--background);
+  flex-shrink: 0;
+  max-height: 4.5em;
+  overflow: hidden;
+}
+
+.tag-pill {
+  font-size: 11px;
+  line-height: 18px;
+  padding: 0 8px;
+  border-radius: 9999px;
+  background-color: color-mix(in srgb, var(--foreground) 8%, transparent);
+  color: var(--foreground);
+  white-space: nowrap;
+  user-select: none;
+}
+
+.dark .tag-pill {
+  background-color: color-mix(in srgb, var(--foreground) 12%, transparent);
+}
+
 .editor-toolbar .separator {
   width: 1px;
   height: 1.5rem;
diff --git a/apps/x/packages/core/src/agents/runtime.ts b/apps/x/packages/core/src/agents/runtime.ts
index 0aeb167f..c84634ac 100644
--- a/apps/x/packages/core/src/agents/runtime.ts
+++ b/apps/x/packages/core/src/agents/runtime.ts
@@ -2,7 +2,6 @@ import { jsonSchema, ModelMessage } from "ai";
 import fs from "fs";
 import path from "path";
 import { WorkDir } from "../config/config.js";
-import { getNoteCreationStrictness } from "../config/note_creation_config.js";
 import { Agent, ToolAttachment } from "@x/shared/dist/agent.js";
 import { AssistantContentPart, AssistantMessage, Message, MessageList, ProviderOptions, ToolCallPart, ToolMessage } from "@x/shared/dist/message.js";
 import { LanguageModel, stepCountIs, streamText, tool, Tool, ToolSet } from "ai";
@@ -25,9 +24,9 @@ import { IRunsLock } from "../runs/lock.js";
 import { IAbortRegistry } from "../runs/abort-registry.js";
 import { PrefixLogger } from "@x/shared";
 import { parse } from "yaml";
-import { raw as noteCreationMediumRaw } from "../knowledge/note_creation_medium.js";
-import { raw as noteCreationLowRaw } from "../knowledge/note_creation_low.js";
-import { raw as noteCreationHighRaw } from "../knowledge/note_creation_high.js";
+import { getRaw as getNoteCreationRaw } from "../knowledge/note_creation.js";
+import { getRaw as getLabelingAgentRaw } from "../knowledge/labeling_agent.js";
+import { getRaw as getNoteTaggingAgentRaw } from "../knowledge/note_tagging_agent.js";
 
 export interface IAgentRuntime {
     trigger(runId: string): Promise<void>;
@@ -316,19 +315,7 @@ export async function loadAgent(id: string): Promise<z.infer<typeof Agent>> {
     }
 
     if (id === 'note_creation') {
-        const strictness = getNoteCreationStrictness();
-        let raw = '';
-        switch (strictness) {
-            case 'medium':
-                raw = noteCreationMediumRaw;
-                break;
-            case 'low':
-                raw = noteCreationLowRaw;
-                break;
-            case 'high':
-                raw = noteCreationHighRaw;
-                break;
-        }
+        const raw = getNoteCreationRaw();
         let agent: z.infer<typeof Agent> = {
             name: id,
             instructions: raw,
@@ -353,6 +340,56 @@ export async function loadAgent(id: string): Promise<z.infer<typeof Agent>> {
         return agent;
     }
 
+    if (id === 'labeling_agent') {
+        const labelingAgentRaw = getLabelingAgentRaw();
+        let agent: z.infer<typeof Agent> = {
+            name: id,
+            instructions: labelingAgentRaw,
+        };
+
+        if (labelingAgentRaw.startsWith("---")) {
+            const end = labelingAgentRaw.indexOf("\n---", 3);
+            if (end !== -1) {
+                const fm = labelingAgentRaw.slice(3, end).trim();
+                const content = labelingAgentRaw.slice(end + 4).trim();
+                const yaml = parse(fm);
+                const parsed = Agent.omit({ name: true, instructions: true }).parse(yaml);
+                agent = {
+                    ...agent,
+                    ...parsed,
+                    instructions: content,
+                };
+            }
+        }
+
+        return agent;
+    }
+
+    if (id === 'note_tagging_agent') {
+        const noteTaggingAgentRaw = getNoteTaggingAgentRaw();
+        let agent: z.infer<typeof Agent> = {
+            name: id,
+            instructions: noteTaggingAgentRaw,
+        };
+
+        if (noteTaggingAgentRaw.startsWith("---")) {
+            const end = noteTaggingAgentRaw.indexOf("\n---", 3);
+            if (end !== -1) {
+                const fm = noteTaggingAgentRaw.slice(3, end).trim();
+                const content = noteTaggingAgentRaw.slice(end + 4).trim();
+                const yaml = parse(fm);
+                const parsed = Agent.omit({ name: true, instructions: true }).parse(yaml);
+                agent = {
+                    ...agent,
+                    ...parsed,
+                    instructions: content,
+                };
+            }
+        }
+
+        return agent;
+    }
+
     const repo = container.resolve<IAgentsRepo>('agentsRepo');
     return await repo.fetch(id);
 }
@@ -706,7 +743,7 @@ export async function* streamAgent({
 
     // set up provider + model
     const provider = createProvider(modelConfig.provider);
-    const knowledgeGraphAgents = ["note_creation", "email-draft", "meeting-prep"];
+    const knowledgeGraphAgents = ["note_creation", "email-draft", "meeting-prep", "labeling_agent", "note_tagging_agent"];
     const modelId = (knowledgeGraphAgents.includes(state.agentName!) && modelConfig.knowledgeGraphModel)
         ? modelConfig.knowledgeGraphModel
         : modelConfig.model;
diff --git a/apps/x/packages/core/src/config/config.ts b/apps/x/packages/core/src/config/config.ts
index 4a91e101..453fef59 100644
--- a/apps/x/packages/core/src/config/config.ts
+++ b/apps/x/packages/core/src/config/config.ts
@@ -23,7 +23,7 @@ function ensureDefaultConfigs() {
     const noteCreationConfig = path.join(WorkDir, "config", "note_creation.json");
     if (!fs.existsSync(noteCreationConfig)) {
         fs.writeFileSync(noteCreationConfig, JSON.stringify({
-            strictness: "high",
+            strictness: "medium",
             configured: false
         }, null, 2));
     }
diff --git a/apps/x/packages/core/src/config/note_creation_config.ts b/apps/x/packages/core/src/config/note_creation_config.ts
index a86e8c00..4aad826f 100644
--- a/apps/x/packages/core/src/config/note_creation_config.ts
+++ b/apps/x/packages/core/src/config/note_creation_config.ts
@@ -11,7 +11,7 @@ interface NoteCreationConfig {
 }
 
 const CONFIG_FILE = path.join(WorkDir, 'config', 'note_creation.json');
-const DEFAULT_STRICTNESS: NoteCreationStrictness = 'high';
+const DEFAULT_STRICTNESS: NoteCreationStrictness = 'medium';
 
 /**
  * Read the full config file.
diff --git a/apps/x/packages/core/src/knowledge/build_graph.ts b/apps/x/packages/core/src/knowledge/build_graph.ts
index a119dfa6..c4174c3f 100644
--- a/apps/x/packages/core/src/knowledge/build_graph.ts
+++ b/apps/x/packages/core/src/knowledge/build_graph.ts
@@ -1,7 +1,6 @@
 import fs from 'fs';
 import path from 'path';
 import { WorkDir } from '../config/config.js';
-import { autoConfigureStrictnessIfNeeded } from '../config/strictness_analyzer.js';
 import { createRun, createMessage } from '../runs/runs.js';
 import { bus } from '../runs/bus.js';
 import { serviceLogger, type ServiceRunContext } from '../services/service_logger.js';
@@ -363,7 +362,19 @@ export async function buildGraph(sourceDir: string): Promise<void> {
     console.log(`[buildGraph] State loaded. Previously processed: ${previouslyProcessedCount} files`);
 
     // Get files that need processing (new or changed)
-    const filesToProcess = getFilesToProcess(sourceDir, state);
+    let filesToProcess = getFilesToProcess(sourceDir, state);
+
+    // For gmail_sync, only process emails that have been labeled (have YAML frontmatter)
+    if (sourceDir.endsWith('gmail_sync')) {
+        filesToProcess = filesToProcess.filter(filePath => {
+            try {
+                const content = fs.readFileSync(filePath, 'utf-8');
+                return content.startsWith('---');
+            } catch {
+                return false;
+            }
+        });
+    }
 
     if (filesToProcess.length === 0) {
         console.log(`[buildGraph] No new or changed files to process in ${path.basename(sourceDir)}`);
@@ -525,8 +536,6 @@ async function processVoiceMemosForKnowledge(): Promise<boolean> {
 async function processAllSources(): Promise<void> {
     console.log('[GraphBuilder] Checking for new content in all sources...');
 
-    // Auto-configure strictness on first run if not already done
-    autoConfigureStrictnessIfNeeded();
 
     let anyFilesProcessed = false;
 
@@ -555,7 +564,19 @@ async function processAllSources(): Promise<void> {
         }
 
         try {
-            const filesToProcess = getFilesToProcess(sourceDir, state);
+            let filesToProcess = getFilesToProcess(sourceDir, state);
+
+            // For gmail_sync, only process emails that have been labeled (have YAML frontmatter)
+            if (folder === 'gmail_sync') {
+                filesToProcess = filesToProcess.filter(filePath => {
+                    try {
+                        const content = fs.readFileSync(filePath, 'utf-8');
+                        return content.startsWith('---');
+                    } catch {
+                        return false;
+                    }
+                });
+            }
 
             if (filesToProcess.length > 0) {
                 console.log(`[GraphBuilder] Found ${filesToProcess.length} new/changed files in ${folder}`);
diff --git a/apps/x/packages/core/src/knowledge/label_emails.ts b/apps/x/packages/core/src/knowledge/label_emails.ts
new file mode 100644
index 00000000..a62f674a
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/label_emails.ts
@@ -0,0 +1,269 @@
+import fs from 'fs';
+import path from 'path';
+import { WorkDir } from '../config/config.js';
+import { createRun, createMessage } from '../runs/runs.js';
+import { bus } from '../runs/bus.js';
+import { serviceLogger } from '../services/service_logger.js';
+import { limitEventItems } from './limit_event_items.js';
+import {
+    loadLabelingState,
+    saveLabelingState,
+    markFileAsLabeled,
+    type LabelingState,
+} from './labeling_state.js';
+
+const SYNC_INTERVAL_MS = 3 * 60 * 1000; // 3 minutes
+const BATCH_SIZE = 15;
+const LABELING_AGENT = 'labeling_agent';
+const GMAIL_SYNC_DIR = path.join(WorkDir, 'gmail_sync');
+const MAX_CONTENT_LENGTH = 8000;
+
+/**
+ * Find email files that haven't been labeled yet
+ */
+function getUnlabeledEmails(state: LabelingState): string[] {
+    if (!fs.existsSync(GMAIL_SYNC_DIR)) {
+        return [];
+    }
+
+    const unlabeled: string[] = [];
+
+    function traverse(dir: string) {
+        const entries = fs.readdirSync(dir);
+        for (const entry of entries) {
+            const fullPath = path.join(dir, entry);
+            const stat = fs.statSync(fullPath);
+
+            if (stat.isDirectory()) {
+                traverse(fullPath);
+            } else if (stat.isFile() && entry.endsWith('.md')) {
+                // Skip if already tracked in state
+                if (state.processedFiles[fullPath]) {
+                    continue;
+                }
+
+                // Skip if file already has frontmatter
+                try {
+                    const content = fs.readFileSync(fullPath, 'utf-8');
+                    if (content.startsWith('---')) {
+                        continue;
+                    }
+                } catch {
+                    continue;
+                }
+
+                unlabeled.push(fullPath);
+            }
+        }
+    }
+
+    traverse(GMAIL_SYNC_DIR);
+    return unlabeled;
+}
+
+/**
+ * Wait for a run to complete by listening for run-processing-end event
+ */
+async function waitForRunCompletion(runId: string): Promise<void> {
+    return new Promise(async (resolve) => {
+        const unsubscribe = await bus.subscribe('*', async (event) => {
+            if (event.type === 'run-processing-end' && event.runId === runId) {
+                unsubscribe();
+                resolve();
+            }
+        });
+    });
+}
+
+/**
+ * Label a batch of email files using the labeling agent
+ */
+async function labelEmailBatch(
+    files: { path: string; content: string }[]
+): Promise<{ runId: string; filesEdited: Set<string> }> {
+    const run = await createRun({
+        agentId: LABELING_AGENT,
+    });
+
+    let message = `Label the following ${files.length} email files by prepending YAML frontmatter.\n\n`;
+    message += `**Important:** Use workspace-relative paths with workspace-edit (e.g. "gmail_sync/email.md", NOT absolute paths).\n\n`;
+
+    for (let i = 0; i < files.length; i++) {
+        const file = files[i];
+        const relativePath = path.relative(WorkDir, file.path);
+        const truncated = file.content.length > MAX_CONTENT_LENGTH
+            ? file.content.slice(0, MAX_CONTENT_LENGTH) + '\n\n[... content truncated, use workspace-readFile for full content ...]'
+            : file.content;
+
+        message += `## File ${i + 1}: ${relativePath}\n\n`;
+        message += truncated;
+        message += `\n\n---\n\n`;
+    }
+
+    const filesEdited = new Set<string>();
+
+    const unsubscribe = await bus.subscribe(run.id, async (event) => {
+        if (event.type !== 'tool-invocation') {
+            return;
+        }
+        if (event.toolName !== 'workspace-edit') {
+            return;
+        }
+        try {
+            const parsed = JSON.parse(event.input) as { path?: string };
+            if (typeof parsed.path === 'string') {
+                filesEdited.add(parsed.path);
+            }
+        } catch {
+            // ignore parse errors
+        }
+    });
+
+    await createMessage(run.id, message);
+    await waitForRunCompletion(run.id);
+    unsubscribe();
+
+    return { runId: run.id, filesEdited };
+}
+
+/**
+ * Process all unlabeled emails in batches
+ */
+async function processUnlabeledEmails(): Promise<void> {
+    console.log('[EmailLabeling] Checking for unlabeled emails...');
+
+    const state = loadLabelingState();
+    const unlabeled = getUnlabeledEmails(state);
+
+    if (unlabeled.length === 0) {
+        console.log('[EmailLabeling] No unlabeled emails found');
+        return;
+    }
+
+    console.log(`[EmailLabeling] Found ${unlabeled.length} unlabeled emails`);
+
+    const run = await serviceLogger.startRun({
+        service: 'email_labeling',
+        message: `Labeling ${unlabeled.length} email${unlabeled.length === 1 ? '' : 's'}`,
+        trigger: 'timer',
+    });
+
+    const relativeFiles = unlabeled.map(f => path.relative(WorkDir, f));
+    const limitedFiles = limitEventItems(relativeFiles);
+    await serviceLogger.log({
+        type: 'changes_identified',
+        service: run.service,
+        runId: run.runId,
+        level: 'info',
+        message: `Found ${unlabeled.length} unlabeled email${unlabeled.length === 1 ? '' : 's'}`,
+        counts: { emails: unlabeled.length },
+        items: limitedFiles.items,
+        truncated: limitedFiles.truncated,
+    });
+
+    const totalBatches = Math.ceil(unlabeled.length / BATCH_SIZE);
+    let totalEdited = 0;
+    let hadError = false;
+
+    for (let i = 0; i < unlabeled.length; i += BATCH_SIZE) {
+        const batchPaths = unlabeled.slice(i, i + BATCH_SIZE);
+        const batchNumber = Math.floor(i / BATCH_SIZE) + 1;
+
+        try {
+            // Read file contents for the batch
+            const files: { path: string; content: string }[] = [];
+            for (const filePath of batchPaths) {
+                try {
+                    const content = fs.readFileSync(filePath, 'utf-8');
+                    files.push({ path: filePath, content });
+                } catch (error) {
+                    console.error(`[EmailLabeling] Error reading ${filePath}:`, error);
+                }
+            }
+
+            if (files.length === 0) {
+                continue;
+            }
+
+            console.log(`[EmailLabeling] Processing batch ${batchNumber}/${totalBatches} (${files.length} files)`);
+            await serviceLogger.log({
+                type: 'progress',
+                service: run.service,
+                runId: run.runId,
+                level: 'info',
+                message: `Processing batch ${batchNumber}/${totalBatches} (${files.length} files)`,
+                step: 'batch',
+                current: batchNumber,
+                total: totalBatches,
+                details: { filesInBatch: files.length },
+            });
+
+            const result = await labelEmailBatch(files);
+            totalEdited += result.filesEdited.size;
+
+            // Only mark files that were actually edited by the agent
+            for (const file of files) {
+                const relativePath = path.relative(WorkDir, file.path);
+                if (result.filesEdited.has(relativePath)) {
+                    markFileAsLabeled(file.path, state);
+                }
+            }
+
+            saveLabelingState(state);
+            console.log(`[EmailLabeling] Batch ${batchNumber}/${totalBatches} complete, ${result.filesEdited.size} files edited`);
+        } catch (error) {
+            hadError = true;
+            console.error(`[EmailLabeling] Error processing batch ${batchNumber}:`, error);
+            await serviceLogger.log({
+                type: 'error',
+                service: run.service,
+                runId: run.runId,
+                level: 'error',
+                message: `Error processing batch ${batchNumber}`,
+                error: error instanceof Error ? error.message : String(error),
+                context: { batchNumber },
+            });
+        }
+    }
+
+    state.lastRunTime = new Date().toISOString();
+    saveLabelingState(state);
+
+    await serviceLogger.log({
+        type: 'run_complete',
+        service: run.service,
+        runId: run.runId,
+        level: hadError ? 'error' : 'info',
+        message: `Email labeling complete: ${totalEdited} files labeled`,
+        durationMs: Date.now() - run.startedAt,
+        outcome: hadError ? 'error' : 'ok',
+        summary: {
+            totalEmails: unlabeled.length,
+            filesLabeled: totalEdited,
+        },
+    });
+
+    console.log(`[EmailLabeling] Done. ${totalEdited} emails labeled.`);
+}
+
+/**
+ * Main entry point - runs as independent polling service
+ */
+export async function init() {
+    console.log('[EmailLabeling] Starting Email Labeling Service...');
+    console.log(`[EmailLabeling] Will check for unlabeled emails every ${SYNC_INTERVAL_MS / 1000} seconds`);
+
+    // Initial run
+    await processUnlabeledEmails();
+
+    // Periodic polling
+    while (true) {
+        await new Promise(resolve => setTimeout(resolve, SYNC_INTERVAL_MS));
+
+        try {
+            await processUnlabeledEmails();
+        } catch (error) {
+            console.error('[EmailLabeling] Error in main loop:', error);
+        }
+    }
+}
diff --git a/apps/x/packages/core/src/knowledge/labeling_agent.ts b/apps/x/packages/core/src/knowledge/labeling_agent.ts
new file mode 100644
index 00000000..f6ff9597
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/labeling_agent.ts
@@ -0,0 +1,59 @@
+import { renderTagSystemForEmails } from './tag_system.js';
+
+export function getRaw(): string {
+  return `---
+model: gpt-5.2
+tools:
+  workspace-readFile:
+    type: builtin
+    name: workspace-readFile
+  workspace-edit:
+    type: builtin
+    name: workspace-edit
+  workspace-readdir:
+    type: builtin
+    name: workspace-readdir
+---
+# Task
+
+You are an email labeling agent. Given a batch of email files, you will classify each email and prepend YAML frontmatter with structured labels.
+
+${renderTagSystemForEmails()}
+
+# Instructions
+
+1. For each email file provided in the message, read its content carefully.
+2. Classify the email using the taxonomy above. Be accurate and conservative — only apply labels that clearly fit.
+3. Use \`workspace-edit\` to prepend YAML frontmatter to the file. The oldString should be the first line of the file (the \`# Subject\` heading), and the newString should be the frontmatter followed by that same first line.
+4. Always include \`processed: true\` and \`labeled_at\` with the current ISO timestamp.
+5. If the email already has frontmatter (starts with \`---\`), skip it.
+
+# Frontmatter Format
+
+\`\`\`yaml
+---
+labels:
+  relationship:
+    - Investor
+  topics:
+    - Fundraising
+    - Finance
+  type: Intro
+  filter:
+    - Promotion
+  action: FYI
+processed: true
+labeled_at: "2026-02-28T12:00:00Z"
+---
+\`\`\`
+
+# Rules
+
+- Every label category must be present in the frontmatter, even if empty (use \`[]\` for empty arrays).
+- \`type\` and \`action\` are single values (strings), not arrays.
+- \`relationship\`, \`topics\`, and \`filter\` are arrays.
+- Use the exact label values from the taxonomy — do not invent new ones.
+- The \`labeled_at\` timestamp should be the current time in ISO 8601 format.
+- Process all files in the batch. Do not skip any unless they already have frontmatter.
+`;
+}
diff --git a/apps/x/packages/core/src/knowledge/labeling_state.ts b/apps/x/packages/core/src/knowledge/labeling_state.ts
new file mode 100644
index 00000000..ced922af
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/labeling_state.ts
@@ -0,0 +1,48 @@
+import fs from 'fs';
+import path from 'path';
+import { WorkDir } from '../config/config.js';
+
+const STATE_FILE = path.join(WorkDir, 'labeling_state.json');
+
+export interface LabelingState {
+    processedFiles: Record<string, { labeledAt: string }>;
+    lastRunTime: string;
+}
+
+export function loadLabelingState(): LabelingState {
+    if (fs.existsSync(STATE_FILE)) {
+        try {
+            return JSON.parse(fs.readFileSync(STATE_FILE, 'utf-8'));
+        } catch (error) {
+            console.error('Error loading labeling state:', error);
+        }
+    }
+
+    return {
+        processedFiles: {},
+        lastRunTime: new Date(0).toISOString(),
+    };
+}
+
+export function saveLabelingState(state: LabelingState): void {
+    try {
+        fs.writeFileSync(STATE_FILE, JSON.stringify(state, null, 2));
+    } catch (error) {
+        console.error('Error saving labeling state:', error);
+        throw error;
+    }
+}
+
+export function markFileAsLabeled(filePath: string, state: LabelingState): void {
+    state.processedFiles[filePath] = {
+        labeledAt: new Date().toISOString(),
+    };
+}
+
+export function resetLabelingState(): void {
+    const emptyState: LabelingState = {
+        processedFiles: {},
+        lastRunTime: new Date().toISOString(),
+    };
+    saveLabelingState(emptyState);
+}
diff --git a/apps/x/packages/core/src/knowledge/note_creation_medium.ts b/apps/x/packages/core/src/knowledge/note_creation.ts
similarity index 70%
rename from apps/x/packages/core/src/knowledge/note_creation_medium.ts
rename to apps/x/packages/core/src/knowledge/note_creation.ts
index 434078c4..d26bd97f 100644
--- a/apps/x/packages/core/src/knowledge/note_creation_medium.ts
+++ b/apps/x/packages/core/src/knowledge/note_creation.ts
@@ -1,4 +1,7 @@
-export const raw = `---
+import { renderNoteTypesBlock } from './note_system.js';
+
+export function getRaw(): string {
+  return `---
 model: gpt-5.2
 tools:
   workspace-writeFile:
@@ -130,25 +133,26 @@ Either:
 
 ---
 
-# The Core Rule: Medium Strictness
+# The Core Rule: Label-Based Filtering
 
-**MEDIUM STRICTNESS MODE**
+**Emails now have YAML frontmatter with labels.** Use these labels to decide whether to process or skip.
 
-**Meetings create notes because:**
-- You chose to spend time with these people
-- If you met them, they matter enough to track
-- Meeting transcripts have rich context
+**Meetings and voice memos always create notes** — no label check needed.
 
-**Emails can create notes if:**
-- The email contains personalized content (not mass mail)
-- The sender seems relevant to your work (business context, not consumer services)
-- The email is part of a meaningful exchange (not one-off transactional)
+**For emails, read the YAML frontmatter labels and apply these rules:**
 
-**Skip creating notes for:**
-- Mass emails and newsletters
-- Automated/transactional emails
-- Consumer service providers (utilities, subscriptions, etc.)
-- Cold sales outreach with no prior relationship indication
+**CREATE/UPDATE notes if the email has ANY of these labels:**
+- **Relationship:** Investor, Customer, Prospect, Partner, Vendor, Candidate, Team, Advisor, Personal, Press, Community, Government
+- **Topic:** Sales, Support, Legal, Finance, Hiring, Fundraising, Event, Research
+- **Type:** Intro, Followup
+- **Action:** Action Required, Urgent, Waiting
+
+**SKIP if the email ONLY has these labels (and none from above):**
+- **Relationship:** Product
+- **Topic:** Travel, Shopping, Health, Learning
+- **Type:** Scheduling, Cold Outreach, Newsletter, Notification
+- **Filter:** Spam, Promotion, Social, Forums
+- **Action:** FYI
 
 ---
 
@@ -217,168 +221,55 @@ Emails containing calendar invites (\`.ics\` attachments or inline calendar data
 
 ---
 
-# Step 1: Source Filtering
+# Step 1: Source Filtering (Label-Based)
 
-## Skip These Sources (Both Meetings and Emails)
+## For Meetings and Voice Memos
+Always process — no filtering needed.
 
-### Mass Emails and Newsletters
+## For Emails — Read YAML Frontmatter
 
-**Indicators:**
-- Sent to a list (To: contains multiple addresses, or undisclosed-recipients)
-- Unsubscribe link in body or footer
-- From a no-reply or marketing address (noreply@, newsletter@, marketing@, hello@)
-- Generic greeting ("Hi there", "Dear subscriber", "Hello!")
-- Promotional language ("Don't miss out", "Limited time", "% off")
-- Mailing list headers (List-Unsubscribe, Mailing-List)
-- Sent via marketing platforms (via sendgrid, via mailchimp, etc.)
+Emails have YAML frontmatter with labels prepended by the labeling agent:
 
-**Action:** SKIP with reason "Newsletter/mass email"
+\`\`\`yaml
+---
+labels:
+  relationship:
+    - Investor
+  topics:
+    - Fundraising
+  type: Intro
+  filter: []
+  action: FYI
+processed: true
+labeled_at: "2026-02-28T12:00:00Z"
+---
+\`\`\`
 
-### Product Updates & Changelogs
+## Decision Rules
 
-**Indicators:**
-- Subject contains: "changelog", "what's new", "product update", "release notes", "v1.x", "new features"
-- Content describes feature releases, bug fixes, or product changes
-- Sent to all users/customers (not personalized to you specifically)
-- From tools/SaaS you use: Cal.com, Notion, Slack, Linear, Figma, etc.
-- No action required from you — purely informational
-- Written in announcement style, not conversational
+Check the labels against the create/skip lists:
 
-**Examples to SKIP:**
-- "Cal.com Changelog v6.1" — product update
-- "What's new in Notion - January 2026" — feature announcement
-- "Introducing new Slack features" — product marketing
-- "Linear Release Notes" — changelog
+**CREATE/UPDATE notes if ANY label matches:**
+- relationship: Investor, Customer, Prospect, Partner, Vendor, Candidate, Team, Advisor, Personal, Press, Community, Government
+- topics: Sales, Support, Legal, Finance, Hiring, Fundraising, Event, Research
+- type: Intro, Followup
+- action: Action Required, Urgent, Waiting
 
-**Action:** SKIP with reason "Product update/changelog"
+**SKIP if labels ONLY match:**
+- relationship: Product
+- topics: Travel, Shopping, Health, Learning
+- type: Scheduling, Cold Outreach, Newsletter, Notification
+- filter: Spam, Promotion, Social, Forums
+- action: FYI
 
-### Cold Outreach / Sales Emails
-
-**THE RULE: If someone emails you offering services and you never responded, SKIP.**
-
-It doesn't matter how personalized, detailed, or relevant the pitch seems. If:
-1. They initiated contact (you didn't reach out first)
-2. They're offering services/products
-3. You never replied or engaged
-
-Then it's cold outreach and should be SKIPPED. Do NOT create notes for cold outreach senders or their organizations.
-
-**EXCEPTION:** If they reference a prior real-world interaction, CREATE a note:
-- "Great meeting you at [conference/event]"
-- "Following up on our conversation at..."
-- "It was nice chatting at [place]"
-- "[Mutual contact] suggested I reach out after we met"
-
-This indicates a real relationship that started offline, not cold outreach.
-
-**Indicators:**
-- Unsolicited contact from someone you've never interacted with
-- Offering services you didn't request (HR, payroll, compliance, bookkeeping, recruiting, dev shops, marketing, etc.)
-- Sales-y language: "wanted to reach out", "thought this might help", "quick question about your..."
-- Mentions your company growth/funding/hiring/tech stack as a hook
-- Attaches "free guides", "case studies", "resources", or "frameworks"
-- Asks for a call/meeting without any prior relationship
-- From domains you've never contacted or met with before
-- No existing note for this person or organization
-- **No reply from the user in the email thread**
-
-**Examples to SKIP:**
-- "Saw you raised funding, wanted to reach out about our services"
-- "Quick question about your bookkeeping/compliance/hiring"
-- "Shared this guide that might help with [your problem]"
-- "Noticed you're scaling, we help startups with..."
-- "Would love 15 minutes to show you how we can help"
-- Detailed pitch about HR/payroll/India expansion services (still cold outreach!)
-- Follow-up emails to previous cold outreach that got no response
-
-**Key distinction:**
-- **You reaching out to a vendor** → worth tracking (you initiated)
-- **You replied to their outreach** → worth tracking (you engaged)
-- **Vendor cold emailing you with no response** → SKIP (no relationship exists)
-
-**IMPORTANT: CC'd people on cold outreach**
-When an email is identified as cold outreach, skip notes for ALL parties involved:
-- The sender (the person doing the outreach)
-- Anyone CC'd on the email (colleagues of the sender, other contacts they're trying to connect)
-- The organization they represent
-
-If someone only appears in your memory as "CC'd on outreach emails from [Sender]", they don't warrant a note — they're just incidentally included in cold outreach, not a real relationship.
-
-**Action:** SKIP with reason "Cold outreach/sales email - no engagement from user"
-
-### Automated/Transactional
-
-**Indicators:**
-- From automated systems (notifications@, alerts@, no-reply@)
-- Password resets, login alerts, shipping notifications
-- Calendar invites without substance
-- Receipts and invoices (unless from key vendor/customer)
-- GitHub/Jira/Slack notifications
-
-**Action:** SKIP with reason "Automated/transactional"
-
-### Low-Signal
-
-**Indicators:**
-- Very short with no substance ("Thanks!", "Sounds good", "Got it")
-- Only contains forwarded message with no commentary
-- Auto-replies ("I'm out of office")
-
-**Action:** SKIP with reason "Low signal"
-
-### Consumer Services (Medium strictness specific)
-
-**Indicators:**
-- From consumer service companies (utilities, streaming, retail)
-- Account management emails
-- Subscription confirmations
-- Delivery notifications
-
-**Action:** SKIP with reason "Consumer service"
-
-### Infrastructure & SaaS Providers
-
-**Skip emails from these types of services:**
-- Domain registrars: GoDaddy, Namecheap, Google Domains, Cloudflare
-- Hosting providers: AWS, Google Cloud, Azure, DigitalOcean, Heroku, Vercel, Netlify
-- Email providers: Google Workspace, Microsoft 365, Zoho
-- Payment processors: Stripe, PayPal, Square, Razorpay
-- Developer tools: GitHub, GitLab, Bitbucket, npm, Docker Hub
-- Analytics: Google Analytics, Mixpanel, Amplitude, Segment
-- Auth providers: Auth0, Okta, Firebase Auth
-- Support platforms: Zendesk, Intercom, Freshdesk
-- HR/Payroll: Gusto, Rippling, Deel, Remote
-
-**Indicators:**
-- Automated system notifications (renewal reminders, usage alerts, security notices)
-- No personalized content from a human
-- From domains like @godaddy.com, @aws.amazon.com, @stripe.com, etc.
-- Templates about account status, billing, or technical alerts
-
-**Action:** SKIP with reason "Infrastructure/SaaS provider notification"
-
-## Email-Specific Processing (Medium Strictness)
-
-For emails, evaluate if the content is personalized and business-relevant:
-
-**Create note if:**
-- The email is personally addressed and substantive
-- The sender appears to be from a business/organization relevant to your work
-- The content discusses work, projects, opportunities, or professional topics
-- It's a warm intro from anyone (not just existing contacts)
-- It's a thoughtful cold outreach that's specific to your work
-
-**Do not create note if:**
-- Clearly mass/templated email
-- Consumer service interaction
-- Generic sales pitch with no personalization
+**Logic:** If even one label falls in the "create" list, process the email. Only skip if ALL labels fall in the "skip" list.
 
 ## Filter Decision Output
 
 If skipping:
 \`\`\`
 SKIP
-Reason: {reason}
+Reason: Labels indicate skip-only categories: {list the labels}
 \`\`\`
 
 If processing, continue to Step 2.
@@ -552,16 +443,16 @@ Resolution Map:
 - "the integration" → "Acme Integration" (same project)
 \`\`\`
 
-## 4b: Apply Source Type Rules (Medium Strictness)
+## 4b: Apply Source Type Rules
 
-**If source_type == "meeting":**
+**If source_type == "meeting" or "voice_memo":**
 - Resolved entities → Update existing notes
 - New entities that pass filters → Create new notes
 
-**If source_type == "email" (MEDIUM STRICTNESS):**
+**If source_type == "email":**
+- The email already passed label-based filtering in Step 1
 - Resolved entities → Update existing notes
-- New entities → Create notes IF the email is personalized and business-relevant
-- New entities from cold sales pitches without personalization → Skip
+- New entities → Create notes (the labels already confirmed this email is worth processing)
 
 ## 4c: Disambiguation Rules
 
@@ -628,39 +519,23 @@ For entities not resolved to existing notes, determine if they warrant new notes
 
 ## People
 
-### Who Gets a Note (Medium Strictness)
+### Who Gets a Note
 
 **CREATE a note for people who are:**
 - External (not @user.domain)
 - Attendees in meetings
-- Email correspondents sending personalized, business-relevant content
+- Email correspondents (emails that reach this step already passed label-based filtering)
 - Decision makers or contacts at customers, prospects, or partners
 - Investors or potential investors
 - Candidates you are interviewing
 - Advisors or mentors
 - Key collaborators
 - Introducers who connect you to valuable contacts
-- Anyone reaching out with a specific, relevant opportunity
 
 **DO NOT create notes for:**
-- Transactional service providers (bank employees, support reps)
-- One-time administrative contacts
 - Large group meeting attendees you didn't interact with
 - Internal colleagues (@user.domain)
 - Assistants handling only logistics
-- Generic role-based contacts
-- Consumer service representatives
-- Generic cold sales outreach with no personalization
-
-### The Relevance Test (Medium Strictness)
-
-Ask: Is this person relevant to my professional work or goals?
-
-- Sarah Chen, VP Engineering evaluating your product → **Yes, create note**
-- James from HSBC who set up your account → **No, skip**
-- Investor reaching out about your company → **Yes, create note**
-- Cold recruiter with a generic pitch → **No, skip**
-- Someone reaching out about a specific opportunity → **Yes, create note**
 
 ### Role Inference
 
@@ -1025,153 +900,18 @@ After writing, verify links go both ways.
 
 ---
 
-# Note Templates
-
-## People
-\`\`\`markdown
-# {Full Name}
-
-## Info
-**Role:** {role, or inferred role with qualifier, or leave blank if truly unknown}
-**Organization:** [[Organizations/{organization}]] or leave blank
-**Email:** {email or leave blank}
-**Aliases:** {comma-separated: first name, nicknames, email}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: Who they are, why you know them, what you're working on together.}
-
-## Connected to
-- [[Organizations/{Organization}]] — works at
-- [[People/{Person}]] — {colleague, introduced by, reports to}
-- [[Projects/{Project}]] — {role}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-\`\`\`
-
-## Organizations
-\`\`\`markdown
-# {Organization Name}
-
-## Info
-**Type:** {company|team|institution|other}
-**Industry:** {industry or leave blank}
-**Relationship:** {customer|prospect|partner|competitor|vendor|other}
-**Domain:** {primary email domain}
-**Aliases:** {comma-separated: short names, abbreviations}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this org is, what your relationship is.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Contacts
-{For transactional contacts who don't get their own notes}
-
-## Projects
-- [[Projects/{Project}]] — {relationship}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-\`\`\`
-
-## Projects
-\`\`\`markdown
-# {Project Name}
-
-## Info
-**Type:** {deal|product|initiative|hiring|other}
-**Status:** {active|planning|on hold|completed|cancelled}
-**Started:** {YYYY-MM-DD or leave blank}
-**Last activity:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this project is, goal, current state.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Organizations
-- [[Organizations/{Org}]] — {customer|partner|etc.}
-
-## Related
-- [[Topics/{Topic}]] — {relationship}
-- [[Projects/{Project}]] — {relationship}
-
-## Timeline
-**{YYYY-MM-DD}** ({meeting|email})
-{What happened.}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}. {Rationale}.
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-\`\`\`
-
-## Topics
-\`\`\`markdown
-# {Topic Name}
-
-## About
-{1-2 sentences: What this topic covers.}
-
-**Keywords:** {comma-separated}
-**Aliases:** {other ways this topic is referenced}
-**First mentioned:** {YYYY-MM-DD}
-**Last mentioned:** {YYYY-MM-DD}
-
-## Related
-- [[People/{Person}]] — {relationship}
-- [[Organizations/{Org}]] — {relationship}
-- [[Projects/{Project}]] — {relationship}
-
-## Log
-**{YYYY-MM-DD}** ({meeting|email}: {title})
-{Summary with [[Folder/Name]] links}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-\`\`\`
+${renderNoteTypesBlock()}
 
 ---
 
-# Summary: Medium Strictness Rules
+# Summary: Label-Based Rules
 
 | Source Type | Creates Notes? | Updates Notes? | Detects State Changes? |
 |-------------|---------------|----------------|------------------------|
 | Meeting | Yes | Yes | Yes |
 | Voice memo | Yes | Yes | Yes |
-| Email (personalized, business-relevant) | Yes | Yes | Yes |
-| Email (mass/automated/consumer) | No (SKIP) | No | No |
-| Email (cold outreach with personalization) | Yes | Yes | Yes |
-| Email (generic cold outreach) | No | No | No |
+| Email (has create label) | Yes | Yes | Yes |
+| Email (only skip labels) | No (SKIP) | No | No |
 
 **Voice memo activity format:** Always include a link to the source voice memo:
 \`\`\`
@@ -1198,7 +938,7 @@ Before completing, verify:
 
 **Source Type:**
 - [ ] Correctly identified as meeting or email
-- [ ] Applied correct medium strictness rules
+- [ ] Applied label-based filtering rules correctly
 
 **Resolution:**
 - [ ] Extracted all name variants from source
@@ -1233,4 +973,5 @@ Before completing, verify:
 - [ ] Dates are YYYY-MM-DD
 - [ ] Bidirectional links are consistent
 - [ ] New notes in correct folders
-`;
\ No newline at end of file
+`;
+}
\ No newline at end of file
diff --git a/apps/x/packages/core/src/knowledge/note_creation_high.ts b/apps/x/packages/core/src/knowledge/note_creation_high.ts
deleted file mode 100644
index ce15c324..00000000
--- a/apps/x/packages/core/src/knowledge/note_creation_high.ts
+++ /dev/null
@@ -1,1950 +0,0 @@
-export const raw = `---
-model: gpt-5.2
-tools:
-  workspace-writeFile:
-    type: builtin
-    name: workspace-writeFile
-  workspace-readFile:
-    type: builtin
-    name: workspace-readFile
-  workspace-edit:
-    type: builtin
-    name: workspace-edit
-  workspace-readdir:
-    type: builtin
-    name: workspace-readdir
-  workspace-mkdir:
-    type: builtin
-    name: workspace-mkdir
-  workspace-grep:
-    type: builtin
-    name: workspace-grep
-  workspace-glob:
-    type: builtin
-    name: workspace-glob
----
-# Task
-
-You are a memory agent. Given a single source file (email, meeting transcript, or voice memo), you will:
-
-1. **Determine source type (meeting or email)**
-2. **Evaluate if the source is worth processing**
-3. **Search for all existing related notes**
-4. **Resolve entities to canonical names**
-5. Identify new entities worth tracking (meetings only)
-6. Extract structured information (decisions, commitments, key facts)
-7. **Detect state changes (status updates, resolved items, role changes)**
-8. Create new notes (meetings only) or update existing notes
-9. **Apply state changes to existing notes**
-
-The core rule: **Meetings and voice memos create notes. Emails enrich them.**
-
-You have full read access to the existing knowledge directory. Use this extensively to:
-- Find existing notes for people, organizations, projects mentioned
-- Resolve ambiguous names (find existing note for "David")
-- Understand existing relationships before updating
-- Avoid creating duplicate notes
-- Maintain consistency with existing content
-- **Detect when new information changes the state of existing notes**
-
-# Inputs
-
-1. **source_file**: Path to a single file to process (email or meeting transcript)
-2. **knowledge_folder**: Path to Obsidian vault (read/write access)
-3. **user**: Information about the owner of this memory
-   - name: e.g., "Arj"
-   - email: e.g., "arj@rowboat.com"
-   - domain: e.g., "rowboat.com"
-4. **knowledge_index**: A pre-built index of all existing notes (provided in the message)
-
-# Knowledge Base Index
-
-**IMPORTANT:** You will receive a pre-built index of all existing notes at the start of each request. This index contains:
-- All people notes with their names, emails, aliases, and organizations
-- All organization notes with their names, domains, and aliases
-- All project notes with their names and statuses
-- All topic notes with their names and keywords
-
-**USE THE INDEX for entity resolution instead of grep/search commands.** This is much faster.
-
-When you need to:
-- Check if a person exists → Look up by name/email/alias in the index
-- Find an organization → Look up by name/domain in the index
-- Resolve "David" to a full name → Check index for people with that name/alias + organization context
-
-**Only use \`cat\` to read full note content** when you need details not in the index (e.g., existing activity logs, open items).
-
-# Tools Available
-
-You have access to these tools:
-
-**For reading files:**
-\`\`\`
-workspace-readFile({ path: "knowledge/People/Sarah Chen.md" })
-\`\`\`
-
-**For creating NEW files:**
-\`\`\`
-workspace-writeFile({ path: "knowledge/People/Sarah Chen.md", data: "# Sarah Chen\\n\\n..." })
-\`\`\`
-
-**For editing EXISTING files (preferred for updates):**
-\`\`\`
-workspace-edit({
-  path: "knowledge/People/Sarah Chen.md",
-  oldString: "## Activity\\n",
-  newString: "## Activity\\n- **2026-02-03** (meeting): New activity entry\\n"
-})
-\`\`\`
-
-**For listing directories:**
-\`\`\`
-workspace-readdir({ path: "knowledge/People" })
-\`\`\`
-
-**For creating directories:**
-\`\`\`
-workspace-mkdir({ path: "knowledge/Projects", recursive: true })
-\`\`\`
-
-**For searching files:**
-\`\`\`
-workspace-grep({ pattern: "Acme Corp", searchPath: "knowledge", fileGlob: "*.md" })
-\`\`\`
-
-**For finding files by pattern:**
-\`\`\`
-workspace-glob({ pattern: "**/*.md", cwd: "knowledge/People" })
-\`\`\`
-
-**IMPORTANT:**
-- Use \`workspace-edit\` for updating existing notes (adding activity, updating fields)
-- Use \`workspace-writeFile\` only for creating new notes
-- Prefer the knowledge_index for entity resolution (it's faster than grep)
-
-# Output
-
-Either:
-- **SKIP** with reason, if source should be ignored
-- Updated or new markdown files in notes_folder
-
----
-
-# The Core Rule: Meetings Create, Emails Enrich
-
-**Meetings create notes because:**
-- You chose to spend time with these people
-- If you met them, they matter enough to track
-- Meeting transcripts have rich context
-
-**Emails only update existing notes because:**
-- Most emails are noise
-- Without a meeting, there's no established relationship worth tracking
-- Prevents memory bloat from random contacts
-
-**The only exception:** Warm intros from someone already in your memory.
-
----
-
-# Step 0: Determine Source Type
-
-Read the source file and determine if it's a meeting or email.
-\`\`\`
-workspace-readFile({ path: "{source_file}" })
-\`\`\`
-
-**Meeting indicators:**
-- Has \`Attendees:\` field
-- Has \`Meeting:\` title
-- Transcript format with speaker labels
-
-**Email indicators:**
-- Has \`From:\` and \`To:\` fields
-- Has \`Subject:\` field
-- Email signature
-
-**Voice memo indicators:**
-- Has \`**Type:** voice memo\` field
-- Has \`**Path:**\` field with path like \`Voice Memos/YYYY-MM-DD/...\`
-- Has \`## Transcript\` section
-
-**Set processing mode:**
-- \`source_type = "meeting"\` → Can create new notes
-- \`source_type = "email"\` → Can only update existing notes
-- \`source_type = "voice_memo"\` → Can create new notes (treat like meetings)
-
----
-
-## Calendar Invite Emails
-
-Emails containing calendar invites (\`.ics\` attachments or inline calendar data) are **high signal** - a scheduled meeting means this person matters.
-
-**How to identify:**
-- Subject contains "Invitation:", "Accepted:", "Declined:", or "Updated:"
-- Has \`.ics\` attachment reference
-- Contains calendar metadata (VCALENDAR, VEVENT)
-
-**Rules for calendar invite emails:**
-1. **CREATE a note for the primary contact** - the person you're actually meeting with
-2. **Extract from the invite:** their name, email, organization (from email domain), meeting topic
-3. **Skip automated notifications from Google/Outlook** - emails from calendar-no-reply@google.com with no human sender
-4. **Skip "Accepted/Declined" responses** - these are just RSVP confirmations, not new contacts
-
-**Who is the primary contact?**
-- For 1:1 meetings: the other person
-- For group meetings: the organizer (unless it's an EA - check if organizer differs from attendees)
-- Look at the meeting title for hints (e.g., "Coffee with Sarah" → Sarah is the contact)
-
-**What to extract:**
-- Name and email from the invite
-- Organization from email domain
-- Meeting topic as context
-- Note that you have an upcoming meeting scheduled
-
-**Examples:**
-- "Invitation: Coffee with Sarah Chen" from sarah@acme.com → CREATE note for Sarah Chen at Acme
-- "Invitation: Acme <> YourCompany sync" organized by sarah@acme.com → CREATE note for Sarah
-- "Accepted: Meeting" from calendar-no-reply@google.com → SKIP (just a notification)
-- "Declined: Sync" from john@example.com → SKIP (RSVP, not a new relationship)
-
-**Why this matters:** Once a note exists, subsequent emails from this person will enrich it. When the meeting happens, the transcript adds more detail.
-
----
-
-# Step 1: Source Filtering
-
-## Skip These Sources (Both Meetings and Emails)
-
-### Mass Emails and Newsletters
-
-**Indicators:**
-- Sent to a list (To: contains multiple addresses, or undisclosed-recipients)
-- Unsubscribe link in body or footer
-- From a no-reply or marketing address (noreply@, newsletter@, marketing@, hello@)
-- Generic greeting ("Hi there", "Dear subscriber", "Hello!")
-- Promotional language ("Don't miss out", "Limited time", "% off")
-- Mailing list headers (List-Unsubscribe, Mailing-List)
-- Sent via marketing platforms (via sendgrid, via mailchimp, etc.)
-
-**Action:** SKIP with reason "Newsletter/mass email"
-
-### Product Updates & Changelogs
-
-**Indicators:**
-- Subject contains: "changelog", "what's new", "product update", "release notes", "v1.x", "new features"
-- Content describes feature releases, bug fixes, or product changes
-- Sent to all users/customers (not personalized to you specifically)
-- From tools/SaaS you use: Cal.com, Notion, Slack, Linear, Figma, etc.
-- No action required from you — purely informational
-- Written in announcement style, not conversational
-
-**Examples to SKIP:**
-- "Cal.com Changelog v6.1" — product update
-- "What's new in Notion - January 2026" — feature announcement
-- "Introducing new Slack features" — product marketing
-- "Linear Release Notes" — changelog
-
-**Action:** SKIP with reason "Product update/changelog"
-
-### Cold Outreach / Sales Emails
-
-**THE RULE: If someone emails you offering services and you never responded, SKIP.**
-
-It doesn't matter how personalized, detailed, or relevant the pitch seems. If:
-1. They initiated contact (you didn't reach out first)
-2. They're offering services/products
-3. You never replied or engaged
-
-Then it's cold outreach and should be SKIPPED. Do NOT create notes for cold outreach senders or their organizations.
-
-**EXCEPTION:** If they reference a prior real-world interaction, CREATE a note:
-- "Great meeting you at [conference/event]"
-- "Following up on our conversation at..."
-- "It was nice chatting at [place]"
-- "[Mutual contact] suggested I reach out after we met"
-
-This indicates a real relationship that started offline, not cold outreach.
-
-**Indicators:**
-- Unsolicited contact from someone you've never interacted with
-- Offering services you didn't request (HR, payroll, compliance, bookkeeping, recruiting, dev shops, marketing, etc.)
-- Sales-y language: "wanted to reach out", "thought this might help", "quick question about your..."
-- Mentions your company growth/funding/hiring/tech stack as a hook
-- Attaches "free guides", "case studies", "resources", or "frameworks"
-- Asks for a call/meeting without any prior relationship
-- From domains you've never contacted or met with before
-- No existing note for this person or organization
-- **No reply from the user in the email thread**
-
-**Examples to SKIP:**
-- "Saw you raised funding, wanted to reach out about our services"
-- "Quick question about your bookkeeping/compliance/hiring"
-- "Shared this guide that might help with [your problem]"
-- "Noticed you're scaling, we help startups with..."
-- "Would love 15 minutes to show you how we can help"
-- Detailed pitch about HR/payroll/India expansion services (still cold outreach!)
-- Follow-up emails to previous cold outreach that got no response
-
-**Key distinction:**
-- **You reaching out to a vendor** → worth tracking (you initiated)
-- **You replied to their outreach** → worth tracking (you engaged)
-- **Vendor cold emailing you with no response** → SKIP (no relationship exists)
-
-**IMPORTANT: CC'd people on cold outreach**
-When an email is identified as cold outreach, skip notes for ALL parties involved:
-- The sender (the person doing the outreach)
-- Anyone CC'd on the email (colleagues of the sender, other contacts they're trying to connect)
-- The organization they represent
-
-If someone only appears in your memory as "CC'd on outreach emails from [Sender]", they don't warrant a note — they're just incidentally included in cold outreach, not a real relationship.
-
-**Action:** SKIP with reason "Cold outreach/sales email - no engagement from user"
-
-### Automated/Transactional
-
-**Indicators:**
-- From automated systems (notifications@, alerts@, no-reply@)
-- Password resets, login alerts, shipping notifications
-- Calendar invites without substance
-- Receipts and invoices (unless from key vendor/customer)
-- GitHub/Jira/Slack notifications
-
-**Action:** SKIP with reason "Automated/transactional"
-
-### Low-Signal
-
-**Indicators:**
-- Very short with no substance ("Thanks!", "Sounds good", "Got it")
-- Only contains forwarded message with no commentary
-- Auto-replies ("I'm out of office")
-
-**Action:** SKIP with reason "Low signal"
-
-### Infrastructure & SaaS Providers
-
-**Skip emails from these types of services:**
-- Domain registrars: GoDaddy, Namecheap, Google Domains, Cloudflare
-- Hosting providers: AWS, Google Cloud, Azure, DigitalOcean, Heroku, Vercel, Netlify
-- Email providers: Google Workspace, Microsoft 365, Zoho
-- Payment processors: Stripe, PayPal, Square, Razorpay
-- Developer tools: GitHub, GitLab, Bitbucket, npm, Docker Hub
-- Analytics: Google Analytics, Mixpanel, Amplitude, Segment
-- Auth providers: Auth0, Okta, Firebase Auth
-- Support platforms: Zendesk, Intercom, Freshdesk
-- HR/Payroll: Gusto, Rippling, Deel, Remote
-
-**Indicators:**
-- Automated system notifications (renewal reminders, usage alerts, security notices)
-- No personalized content from a human
-- From domains like @godaddy.com, @aws.amazon.com, @stripe.com, etc.
-- Templates about account status, billing, or technical alerts
-
-**Action:** SKIP with reason "Infrastructure/SaaS provider notification"
-
-## Email-Specific Filtering
-
-For emails, check if sender/recipients have existing notes:
-\`\`\`
-workspace-grep({ pattern: "{sender email}", searchPath: "{knowledge_folder}" })
-workspace-grep({ pattern: "{sender name}", searchPath: "{knowledge_folder}/People" })
-\`\`\`
-
-**If no existing note found:**
-- Check if this is a warm intro from someone in memory (see below)
-- If not a warm intro → SKIP with reason "No existing relationship"
-
-**If existing note found:**
-- Continue processing
-- Will update existing note only
-
-### Detecting Warm Intros
-
-A warm intro is when someone already in your memory introduces you to someone new.
-
-**Indicators:**
-- Subject contains "Intro:" or "Introduction:"
-- Body contains "want to introduce" or "meet [Name]"
-- Sender has an existing note in memory
-- New person is CC'd or mentioned
-
-**If warm intro detected:**
-- This is the ONE exception where email can create notes
-- Create note for the introduced person
-- Create org note for their company if needed
-
-## Filter Decision Output
-
-If skipping:
-\`\`\`
-SKIP
-Reason: {reason}
-\`\`\`
-
-If processing, continue to Step 2.
-
----
-
-# Step 2: Read and Parse Source File
-\`\`\`
-workspace-readFile({ path: "{source_file}" })
-\`\`\`
-
-Extract metadata:
-
-**For meetings:**
-- **Date:** From header or filename
-- **Title:** Meeting name
-- **Attendees:** List of participants
-- **Duration:** If available
-
-**For emails:**
-- **Date:** From \`Date:\` header
-- **Subject:** From \`Subject:\` header
-- **From:** Sender email/name
-- **To/Cc:** Recipients
-
-## 2a: Exclude Self
-
-Never create or update notes for:
-- The user (matches user.name, user.email, or @user.domain)
-- Anyone @{user.domain} (colleagues at user's company)
-
-Filter these out from attendees/participants before proceeding.
-
-## 2b: Extract All Name Variants
-
-From the source, collect every way entities are referenced:
-
-**People variants:**
-- Full names: "Sarah Chen"
-- First names only: "Sarah"
-- Last names only: "Chen"
-- Initials: "S. Chen"
-- Email addresses: "sarah@acme.com"
-- Roles/titles: "their CTO", "the VP of Engineering"
-- Pronouns with clear antecedents: "she" (referring to Sarah in same paragraph)
-
-**Organization variants:**
-- Full names: "Acme Corporation"
-- Short names: "Acme"
-- Abbreviations: "AC"
-- Email domains: "@acme.com"
-- References: "your company", "their team"
-
-**Project variants:**
-- Explicit names: "Project Atlas"
-- Descriptive references: "the integration", "the pilot", "the deal"
-- Combined references: "Acme integration", "the Series A"
-
-Create a list of all variants found:
-\`\`\`
-Variants found:
-- People: "Sarah Chen", "Sarah", "sarah@acme.com", "David", "their CTO"
-- Organizations: "Acme Corp", "Acme", "@acme.com"
-- Projects: "the pilot", "Q2 integration"
-\`\`\`
-
----
-
-# Step 3: Look Up Existing Notes in Index
-
-**Use the provided knowledge_index to find existing notes. Do NOT use grep commands.**
-
-## 3a: Look Up People
-
-For each person variant (name, email, alias), check the index:
-
-\`\`\`
-From index, find matches for:
-- "Sarah Chen" → Check People table for matching name
-- "Sarah" → Check People table for matching name or alias
-- "sarah@acme.com" → Check People table for matching email
-- "@acme.com" → Check People table for matching organization or check Organizations for domain
-\`\`\`
-
-## 3b: Look Up Organizations
-
-\`\`\`
-From index, find matches for:
-- "Acme Corp" → Check Organizations table for matching name
-- "Acme" → Check Organizations table for matching name or alias
-- "acme.com" → Check Organizations table for matching domain
-\`\`\`
-
-## 3c: Look Up Projects and Topics
-
-\`\`\`
-From index, find matches for:
-- "the pilot" → Check Projects table for related names
-- "SOC 2" → Check Topics table for matching keywords
-\`\`\`
-
-## 3d: Read Full Notes When Needed
-
-Only read the full note content when you need details not in the index (e.g., activity logs, open items):
-\`\`\`bash
-workspace-readFile({ path: "{knowledge_folder}/People/Sarah Chen.md" })
-\`\`\`
-
-**Why read these notes:**
-- Find canonical names (David → David Kim)
-- Check Aliases fields for known variants
-- Understand existing relationships
-- See organization context for disambiguation
-- Check what's already captured (avoid duplicates)
-- Review open items (some might be resolved)
-- **Check current status fields (might need updating)**
-- **Check current roles (might have changed)**
-
-## 3e: Matching Criteria
-
-Use these criteria to determine if a variant matches an existing note:
-
-**People matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| First name "Sarah" | Full name "Sarah Chen" | Same organization context |
-| Email "sarah@acme.com" | Email field | Exact match |
-| Email domain "@acme.com" | Organization "Acme Corp" | Domain matches org |
-| Role "VP Engineering" | Role field | Same org + same role |
-| First name + company context | Full name + Organization | Company matches |
-| Any variant | Aliases field | Listed in aliases |
-
-**Organization matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| "Acme" | "Acme Corp" | Substring match |
-| "Acme Corporation" | "Acme Corp" | Same root name |
-| "@acme.com" | Domain field | Domain matches |
-| Any variant | Aliases field | Listed in aliases |
-
-**Project matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| "the pilot" | "Acme Pilot" | Same org context in source |
-| "integration project" | "Acme Integration" | Same org + similar type |
-| "Series A" | "Series A Fundraise" | Unique identifier match |
-
----
-
-# Step 4: Resolve Entities to Canonical Names
-
-Using the search results from Step 3, resolve each variant to a canonical name.
-
-## 4a: Build Resolution Map
-
-Create a mapping from every source reference to its canonical form:
-\`\`\`
-Resolution Map:
-- "Sarah Chen" → "Sarah Chen" (exact match found)
-- "Sarah" → "Sarah Chen" (matched via Acme context)
-- "sarah@acme.com" → "Sarah Chen" (email match in note)
-- "David" → "David Kim" (matched via Acme context)
-- "their CTO" → "Jennifer Lee" (role match at Acme) OR "Unknown CTO at Acme Corp" (if not found)
-- "Acme" → "Acme Corp" (existing note)
-- "Acme Corporation" → "Acme Corp" (alias match)
-- "@acme.com" → "Acme Corp" (domain match)
-- "the pilot" → "Acme Integration" (project with Acme)
-- "the integration" → "Acme Integration" (same project)
-\`\`\`
-
-## 4b: Apply Source Type Rules
-
-**If source_type == "meeting":**
-- Resolved entities → Update existing notes
-- New entities that pass filters → Create new notes
-
-**If source_type == "email":**
-- Resolved entities → Update existing notes
-- New entities → Do NOT create notes (skip them)
-- Exception: Warm intro → Create note for introduced person
-
-## 4c: Disambiguation Rules
-
-When multiple candidates match a variant, disambiguate:
-
-**By organization (strongest signal):**
-\`\`\`
-# "David" could be David Kim or David Chen
-workspace-grep({ pattern: "Acme", searchPath: "{knowledge_folder}/People/David Kim.md" })
-# Output: **Organization:** [[Acme Corp]]
-
-workspace-grep({ pattern: "Acme", searchPath: "{knowledge_folder}/People/David Chen.md" })
-# Output: **Organization:** [[Other Corp]]
-
-# Source is from Acme context → "David" = "David Kim"
-\`\`\`
-
-**By email (definitive):**
-\`\`\`
-workspace-grep({ pattern: "david@acme.com", searchPath: "{knowledge_folder}/People/David Kim.md" })
-# Exact email match is definitive
-\`\`\`
-
-**By role:**
-\`\`\`
-# Source mentions "their CTO"
-workspace-grep({ pattern: "Role.*CTO", searchPath: "{knowledge_folder}/People" })
-# Filter results by organization context
-\`\`\`
-
-**By recency (weakest signal):**
-If still ambiguous, prefer the person with more recent activity in notes.
-
-**If still ambiguous:**
-- Flag in resolution map: "David" → "David (ambiguous - could be David Kim or David Chen)"
-- Will handle in Step 5
-
-## 4d: Resolution Map Output
-
-Final resolution map before proceeding:
-\`\`\`
-RESOLVED (use canonical name with absolute path):
-- "Sarah", "Sarah Chen", "sarah@acme.com" → [[People/Sarah Chen]]
-- "David" → [[People/David Kim]]
-- "Acme", "Acme Corp", "@acme.com" → [[Organizations/Acme Corp]]
-- "the pilot", "the integration" → [[Projects/Acme Integration]]
-
-NEW ENTITIES (meetings only — create notes):
-- "Jennifer" (CTO, Acme Corp) → Create [[People/Jennifer]] or [[People/Jennifer (Acme Corp)]]
-- "SOC 2" → Create [[Topics/Security Compliance]]
-
-NEW ENTITIES (emails — do not create):
-- "Random Person" → Skip, no existing relationship
-
-AMBIGUOUS (flag or skip):
-- "Mike" (no context) → Mention in activity only, don't create note
-
-SKIP (doesn't warrant note):
-- "their assistant" → Transactional contact
-\`\`\`
-
----
-
-# Step 5: Identify New Entities (Meetings Only)
-
-**This step only applies to meetings. For emails, skip to Step 6.**
-
-For entities not resolved to existing notes, determine if they warrant new notes.
-
-## People (Meetings Only)
-
-### Who Gets a Note
-
-**CREATE a note for meeting attendees who are:**
-- External (not @user.domain)
-- Decision makers or key contacts at customers, prospects, or partners
-- Investors or potential investors
-- Candidates you are interviewing
-- Advisors or mentors with ongoing relationships
-- Key collaborators on important matters
-- Introducers who connect you to valuable contacts
-
-**DO NOT create notes for:**
-- Transactional service providers (bank employees, support reps)
-- One-time administrative contacts
-- Large group meeting attendees you didn't interact with
-- Internal colleagues (@user.domain)
-- Assistants handling only logistics
-- Generic role-based contacts
-
-### The "Would I Prep for This Person?" Test
-
-Ask: If I had a call with this person next week, would I want notes beforehand?
-
-- Sarah Chen, VP Engineering evaluating your product → **Yes, create note**
-- James from HSBC who set up your account → **No, skip**
-- Investor you're pitching → **Yes, create note**
-- Recruiter scheduling interviews → **No, skip**
-
-### Role Inference
-
-If role is not explicitly stated, infer from context:
-
-**From email signatures:**
-- Often contains title
-
-**From meeting context:**
-- Organizer of cross-company meeting → likely senior or partnerships
-- Technical questions → likely engineering
-- Pricing questions → likely procurement or finance
-- Product feedback → likely product
-
-**From email patterns:**
-- firstname@company.com → often founder or senior
-- firstname.lastname@company.com → often larger company employee
-
-**From conversation content:**
-- "I'll need to check with my team" → manager
-- "Let me run this by leadership" → IC or mid-level
-- "I can make that call" → decision maker
-
-**Format in note:**
-\`\`\`markdown
-**Role:** Product Lead (inferred from evaluation discussions)
-**Role:** Senior (inferred — organized cross-company meeting)
-**Role:** Engineering (inferred — asked technical integration questions)
-\`\`\`
-
-**Never write just "Unknown" if you can make a reasonable inference.**
-
-### Relationship Type Guide
-
-| Relationship Type | Create People Notes? | Create Org Note? |
-|-------------------|----------------------|------------------|
-| Customer (active deal) | Yes — key contacts | Yes |
-| Customer (support ticket) | No | Maybe update existing |
-| Prospect | Yes — decision makers | Yes |
-| Investor | Yes | Yes |
-| Strategic partner | Yes — key contacts | Yes |
-| Vendor (strategic) | Yes — main contact only | Yes |
-| Vendor (transactional) | No | Optional |
-| Bank/Financial services | No | Yes (one note) |
-| Candidate | Yes | No |
-| Service provider (one-time) | No | No |
-
-### Handling Non-Note-Worthy People
-
-For people who don't warrant their own note, add to Organization note's Contacts section:
-\`\`\`markdown
-## Contacts
-- James Wong — Relationship Manager, helped with account setup
-- Sarah Lee — Support, handled wire transfer issue
-\`\`\`
-
-## Organizations (Meetings Only)
-
-**CREATE a note if:**
-- Someone from that org attended the meeting
-- It's a customer, prospect, investor, or partner
-
-**DO NOT create for:**
-- Tool/service providers mentioned in passing
-- One-time transactional vendors
-
-## Projects (Meetings Only)
-
-**CREATE a note if:**
-- Discussed substantively in the meeting
-- Has a goal and timeline
-- Involves multiple interactions
-
-## Topics (Meetings Only)
-
-**CREATE a note if:**
-- Recurring theme discussed
-- Will come up again across conversations
-
----
-
-# Step 6: Extract Content
-
-For each entity that has or will have a note, extract relevant content.
-
-## Decisions
-
-**Indicators:**
-- "We decided..." / "We agreed..." / "Let's go with..."
-- "The plan is..." / "Going forward..."
-- "Approved" / "Confirmed" / "Chose X over Y"
-
-**Extract:** What, when (source date), who, rationale.
-
-## Commitments
-
-**Indicators:**
-- "I'll..." / "We'll..." / "Let me..."
-- "Can you..." / "Please send..."
-- "By Friday" / "Next week" / "Before the call"
-
-**Extract:** Owner, action, deadline, status (open).
-
-## Key Facts
-
-Key facts should be **substantive information about the entity** — not commentary about missing data.
-
-**Extract if:**
-- Specific numbers (budget: $50K, team size: 12, timeline: Q2)
-- Preferences or working style ("prefers async communication")
-- Background information ("previously at Google")
-- Authority or decision process ("needs CEO sign-off")
-- Concerns or constraints ("security is top priority")
-- What they're evaluating or interested in
-- What was discussed or proposed
-- Technical requirements or specifications
-
-**Never include:**
-- Meta-commentary about missing data ("Name only provided", "Role not mentioned")
-- Obvious facts ("Works at Acme" — that's in the Info section)
-- Placeholder text ("Unknown", "TBD")
-- Data quality observations ("Full name not in email")
-
-**If there are no substantive key facts, leave the section empty.** An empty section is better than filler.
-
-**Good key facts:**
-\`\`\`markdown
-## Key facts
-- Evaluating AI copilot for in-app experience
-- Three use cases discussed: pre-purchase sales, onboarding, coaching
-- Budget approved for Q2 pilot
-- Needs SOC 2 compliance before proceeding
-\`\`\`
-
-**Bad key facts:**
-\`\`\`markdown
-## Key facts
-- Name only provided; full name/role not in email.
-- Email address not available.
-- Meeting was 50 minutes.
-\`\`\`
-
-## Open Items
-
-Open items are **commitments and next steps from the conversation** — not tasks to fill in missing data.
-
-**Include:**
-- Commitments made: "I'll send the documentation by Friday"
-- Requests received: "Can you share pricing?"
-- Next steps discussed: "Let's schedule a technical deep-dive"
-- Follow-ups agreed: "Will loop in their CTO"
-
-**Format:**
-\`\`\`markdown
-- [ ] {Action} — {owner if not you}, {due date if known}
-\`\`\`
-
-**Never include:**
-- Data gaps: "Find their full name", "Get their email", "Add role"
-- Wishes: "Would be good to know their budget"
-- Agent tasks: "Research their company"
-
-**If there are no actual commitments or next steps, leave the section empty.**
-
-**Good open items:**
-\`\`\`markdown
-## Open items
-- [ ] Send API documentation — by Friday
-- [ ] Schedule follow-up call with CTO
-- [ ] Share pricing proposal — after technical review
-\`\`\`
-
-**Bad open items:**
-\`\`\`markdown
-## Open items
-- [ ] Find Matteo's full name, role, and email at [[Eight Sleep]]
-- [ ] Add Anurag's role/title at Groww
-- [ ] Research Eight Sleep company background
-\`\`\`
-
-## Summary
-
-The summary should answer: **"Who is this person and why do I know them?"**
-
-**Write 2-3 sentences covering:**
-- Their role/function (even if inferred)
-- The context of your relationship
-- What you're discussing or working on together
-
-**Focus on the relationship, not the communication method.**
-
-**Good summaries:**
-\`\`\`markdown
-## Summary
-Product contact at [[Organizations/Eight Sleep]] exploring an AI copilot for their app.
-Initial discussions covered sales assistance, onboarding, and coaching use cases.
-Currently evaluating fit with their product roadmap.
-\`\`\`
-\`\`\`markdown
-## Summary
-VP Engineering at [[Organizations/Acme Corp]] leading their integration project.
-Key technical decision-maker. Working toward Q2 pilot launch.
-\`\`\`
-
-**Bad summaries:**
-\`\`\`markdown
-## Summary
-Contact at [[Organizations/Eight Sleep]]; received an outbound pitch from [[People/Arjun Maheswaran]]
-about an in-app AI copilot concept.
-\`\`\`
-\`\`\`markdown
-## Summary
-Attendee on the scheduled "Groww <> RowBoat" meeting (Aug 12, 2024).
-\`\`\`
-
-**Why these are bad:**
-- "Received an outbound pitch" — describes the email, not the relationship
-- "Attendee on scheduled meeting" — describes attendance, not who they are
-
-**Infer when needed:**
-If role is unknown but context suggests it, say so:
-- "Likely product or partnerships (evaluating AI integration)"
-- "Senior contact (organized cross-company meeting)"
-
-## Activity Summary
-
-One line summarizing this source's relevance to the entity:
-\`\`\`
-**{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[links]]}
-\`\`\`
-
-**For voice memos:** Include a link to the voice memo file using the Path field:
-\`\`\`
-**2025-01-15** (voice memo): Discussed [[Projects/Acme Integration]] timeline. See [[Voice Memos/2025-01-15/voice-memo-2025-01-15T10-30-00-000Z]]
-\`\`\`
-
-**Important:** Use canonical names with absolute paths from resolution map in all summaries:
-\`\`\`
-# Correct (uses absolute paths):
-**2025-01-15** (meeting): [[People/Sarah Chen]] confirmed timeline with [[People/David Kim]]. Blocked on [[Topics/Security Compliance]].
-
-# Incorrect (uses variants or relative links):
-**2025-01-15** (meeting): Sarah confirmed timeline with David. Blocked on SOC 2.
-**2025-01-15** (meeting): [[Sarah Chen]] confirmed timeline with [[David Kim]]. Blocked on [[Security Compliance]].
-\`\`\`
-
----
-
-# Step 7: Detect State Changes
-
-Review the extracted content for signals that existing note fields should be updated.
-
-## 7a: Project Status Changes
-
-**Look for these signals:**
-
-| Signal | New Status |
-|--------|------------|
-| "Moving forward" / "approved" / "signed" / "green light" | active |
-| "On hold" / "pausing" / "delayed" / "pushed back" | on hold |
-| "Cancelled" / "not proceeding" / "killed" / "passed" | cancelled |
-| "Launched" / "completed" / "done" / "shipped" | completed |
-| "Exploring" / "considering" / "evaluating" / "might" | planning |
-
-**Action:** If a related project note exists and the signal is clear, update the \`**Status:**\` field.
-
-**Example:**
-\`\`\`
-Source: "Great news — leadership approved the pilot!"
-Current: **Status:** planning
-Update to: **Status:** active
-\`\`\`
-
-**Be conservative:** Only update status when the signal is unambiguous. If unclear, add to activity log but don't change status.
-
-## 7b: Open Item Resolution
-
-**Look for signals that a previously tracked open item is now complete:**
-
-| Signal | Action |
-|--------|--------|
-| "Here's the [X] you requested" | Mark [X] complete |
-| "I've sent the [X]" | Mark [X] complete |
-| "The [X] is ready" | Mark [X] complete |
-| "[X] is done" | Mark [X] complete |
-| "Attached is the [X]" | Mark [X] complete |
-
-**How to match:**
-1. Read existing open items from the note
-2. Look for items that match what was delivered/completed
-3. Change \`- [ ]\` to \`- [x]\` with completion date
-
-**Example:**
-\`\`\`
-Source: "Here's the API documentation you requested."
-Current: - [ ] Send API documentation — by Friday
-Update to: - [x] Send API documentation — completed 2025-01-16
-\`\`\`
-
-**Be conservative:** Only mark complete if there's a clear match. If unsure, add to activity log but don't mark complete.
-
-## 7c: Role/Title Changes
-
-**Look for signals:**
-- New title in email signature
-- "I've been promoted to..."
-- "I'm now the..."
-- "I've moved to the [X] team"
-- Different role mentioned than what's in the note
-
-**Action:** Update the \`**Role:**\` field in person note.
-
-**Example:**
-\`\`\`
-Source: Email signature shows "VP Engineering"
-Current: **Role:** Engineering Lead
-Update to: **Role:** VP Engineering (updated 2025-01-16)
-\`\`\`
-
-## 7d: Organization/Relationship Changes
-
-**Look for signals:**
-- "I've joined [New Company]"
-- "We're now a customer" / "We signed the contract"
-- "We've partnered with..."
-- "They acquired us"
-- New email domain for known person
-
-**Action:** Update relevant fields:
-- Person's \`**Organization:**\` field
-- Org's \`**Relationship:**\` field (prospect → customer, etc.)
-
-**Example:**
-\`\`\`
-Source: "Excited to announce we've signed the contract!"
-Current: **Relationship:** prospect
-Update to: **Relationship:** customer
-\`\`\`
-
-## 7e: Build State Change List
-
-Before writing, compile all detected state changes:
-\`\`\`
-STATE CHANGES:
-- [[Projects/Acme Integration]]: Status planning → active (leadership approved)
-- [[People/Sarah Chen]]: Role "Engineering Lead" → "VP Engineering" (signature)
-- [[People/Sarah Chen]]: Open item "Send API documentation" → completed
-- [[Organizations/Acme Corp]]: Relationship prospect → customer (contract signed)
-\`\`\`
-
----
-
-# Step 8: Check for Duplicates and Conflicts
-
-Before writing, compare extracted content against existing notes.
-
-## Check Activity Log
-\`\`\`
-workspace-grep({ pattern: "2025-01-15", searchPath: "{knowledge_folder}/People/Sarah Chen.md" })
-\`\`\`
-
-If an entry for this date/source already exists, this may have been processed. Skip or verify different interaction.
-
-## Check Key Facts
-
-Review key facts against existing. Skip duplicates.
-
-## Check Open Items
-
-Review open items for:
-- Duplicates (don't add same item twice)
-- Items that should be marked complete (from Step 7b)
-
-## Check for Conflicts
-
-If new info contradicts existing:
-- Note both versions
-- Add "(needs clarification)"
-- Don't silently overwrite
-
----
-
-# Step 9: Write Updates
-
-## 9a: Meetings — Create and Update Notes
-
-**IMPORTANT: Write sequentially, one file at a time.**
-- Generate content for exactly one note.
-- Issue exactly one write/edit command.
-- Wait for the tool to return before generating the next note.
-- Do NOT batch multiple write commands in a single response.
-
-**For NEW entities (use workspace-writeFile):**
-\`\`\`
-workspace-writeFile({
-  path: "{knowledge_folder}/People/Jennifer.md",
-  data: "# Jennifer\\n\\n## Summary\\n..."
-})
-\`\`\`
-
-**For EXISTING entities (use workspace-edit):**
-- Read current content first with workspace-readFile
-- Use workspace-edit to add activity entry at TOP (reverse chronological)
-- Update fields using targeted edits
-\`\`\`
-workspace-edit({
-  path: "{knowledge_folder}/People/Sarah Chen.md",
-  oldString: "## Activity\\n",
-  newString: "## Activity\\n- **2026-02-03** (meeting): Met to discuss project timeline\\n"
-})
-\`\`\`
-
-## 9b: Emails — Update Existing Notes Only
-
-**Only update notes that already exist.**
-
-Do NOT create new notes from emails (except warm intros).
-
-For existing notes:
-- Add activity entry
-- Update "Last seen" date
-- Add new key facts
-- Add new commitments
-- Update open items if resolved
-
-## 9c: Apply State Changes
-
-For each state change identified in Step 7:
-
-### Update Project Status
-\`\`\`bash
-# Read current project note
-workspace-readFile({ path: "{knowledge_folder}/Projects/Acme Integration.md" })
-
-# Update the Status field
-# Change: **Status:** planning
-# To: **Status:** active
-\`\`\`
-
-### Mark Open Items Complete
-\`\`\`bash
-# Read current note
-workspace-readFile({ path: "{knowledge_folder}/People/Sarah Chen.md" })
-
-# Find matching open item and update
-# Change: - [ ] Send API documentation — by Friday
-# To: - [x] Send API documentation — completed 2025-01-16
-\`\`\`
-
-### Update Role
-\`\`\`bash
-# Read current person note
-workspace-readFile({ path: "{knowledge_folder}/People/Sarah Chen.md" })
-
-# Update role field
-# Change: **Role:** Engineering Lead
-# To: **Role:** VP Engineering
-\`\`\`
-
-### Update Relationship
-\`\`\`bash
-# Read current org note
-workspace-readFile({ path: "{knowledge_folder}/Organizations/Acme Corp.md" })
-
-# Update relationship field
-# Change: **Relationship:** prospect
-# To: **Relationship:** customer
-\`\`\`
-
-### Log State Changes in Activity
-
-When applying a state change, also note it in the activity log:
-\`\`\`markdown
-- **2025-01-16** (email): Leadership approved pilot. [Status → active] Contract being drafted.
-\`\`\`
-
-Use \`[Field → new value]\` notation to make state changes visible in the activity log.
-
-## 9d: Update Aliases
-
-If you discovered new name variants during resolution, add them to Aliases field:
-\`\`\`markdown
-# Before
-**Aliases:** Sarah, S. Chen
-
-# Source used "Sarah C." (new variant)
-
-# After  
-**Aliases:** Sarah, S. Chen, Sarah C.
-\`\`\`
-
-## 9e: Writing Rules
-
-- **Always use absolute paths** with format \`[[Folder/Name]]\` for all links
-- Use YYYY-MM-DD format for dates
-- Be concise: one line per activity entry
-- Note state changes with \`[Field → value]\` in activity
-- Escape quotes properly in shell commands
-- Write only one file per response (no multi-file write batches)
-
----
-
-# Step 10: Ensure Bidirectional Links
-
-After writing, verify links go both ways.
-
-## Absolute Link Format
-
-**IMPORTANT:** Always use absolute links with the folder path to avoid ambiguity:
-
-\`\`\`markdown
-[[People/Sarah Chen]]
-[[Organizations/Acme Corp]]
-[[Projects/Acme Integration]]
-[[Topics/Security Compliance]]
-\`\`\`
-
-Format: \`[[Folder/Note Name]]\`
-
-This ensures:
-- No ambiguity when names overlap across folders
-- Clear navigation in any Obsidian-compatible tool
-- Consistent linking throughout the vault
-
-## Check Each New Link
-
-If you added \`[[People/Jennifer]]\` to \`Organizations/Acme Corp.md\`:
-\`\`\`
-workspace-grep({ pattern: "Acme Corp", searchPath: "{knowledge_folder}/People/Jennifer.md" })
-\`\`\`
-
-If not found, update Jennifer.md to add the link.
-
-## Bidirectional Link Rules
-
-| If you add... | Then also add... |
-|---------------|------------------|
-| Person → Organization | Organization → Person (in People section) |
-| Person → Project | Project → Person (in People section) |
-| Project → Organization | Organization → Project (in Projects section) |
-| Project → Topic | Topic → Project (in Related section) |
-| Person → Person | Person → Person (reverse link) |
-
----
-
-# Note Templates
-
-## People
-\`\`\`markdown
-# {Full Name}
-
-## Info
-**Role:** {role, or inferred role with qualifier, or leave blank if truly unknown}
-**Organization:** [[Organizations/{organization}]] or leave blank
-**Email:** {email or leave blank}
-**Aliases:** {comma-separated: first name, nicknames, email}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: Who they are, why you know them, what you're working on together. Focus on relationship and context, not communication method.}
-
-## Connected to
-- [[Organizations/{Organization}]] — works at
-- [[People/{Person}]] — {colleague, introduced by, reports to}
-- [[Projects/{Project}]] — {role}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links} {[State changes if any]}
-
-## Key facts
-{Substantive facts only. Leave empty if none. Never include data gap commentary.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none. Never include "find their email" type items.}
-{Mark completed items with [x] and completion date.}
-\`\`\`
-
-## Organizations
-\`\`\`markdown
-# {Organization Name}
-
-## Info
-**Type:** {company|team|institution|other}
-**Industry:** {industry or leave blank}
-**Relationship:** {customer|prospect|partner|competitor|vendor|other}
-**Domain:** {primary email domain}
-**Aliases:** {comma-separated: short names, abbreviations}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this org is, what your relationship is, what you're working on together.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Contacts
-{For transactional contacts who don't get their own notes}
-- {Name} — {role}, {context}
-
-## Projects
-- [[Projects/{Project}]] — {relationship}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links} {[State changes if any]}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-\`\`\`
-
-## Projects
-\`\`\`markdown
-# {Project Name}
-
-## Info
-**Type:** {deal|product|initiative|hiring|other}
-**Status:** {active|planning|on hold|completed|cancelled}
-**Started:** {YYYY-MM-DD or leave blank}
-**Last activity:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this project is, goal, current state.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Organizations
-- [[Organizations/{Org}]] — {customer|partner|etc.}
-
-## Related
-- [[Topics/{Topic}]] — {relationship}
-- [[Projects/{Project}]] — {relationship}
-
-## Timeline
-**{YYYY-MM-DD}** ({meeting|email})
-{What happened. Key points with [[Folder/Name]] links.} {[Status → new status] if changed}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}. {Rationale}.
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-\`\`\`
-
-## Topics
-\`\`\`markdown
-# {Topic Name}
-
-## About
-{1-2 sentences: What this topic covers.}
-
-**Keywords:** {comma-separated}
-**Aliases:** {other ways this topic is referenced}
-**First mentioned:** {YYYY-MM-DD}
-**Last mentioned:** {YYYY-MM-DD}
-
-## Related
-- [[People/{Person}]] — {relationship}
-- [[Organizations/{Org}]] — {relationship}
-- [[Projects/{Project}]] — {relationship}
-
-## Log
-**{YYYY-MM-DD}** ({meeting|email}: {title})
-{Summary with [[Folder/Name]] links}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-\`\`\`
-
----
-
-# Named Entity Resolution Reference
-
-## Quick Algorithm
-
-1. Extract all name variants from source
-2. Search notes folder for each variant (including Aliases fields)
-3. Read candidate notes, check org/role/email context
-4. Disambiguate: org context > email match > role match > recency
-5. Build resolution map
-6. Apply source type rules (meetings create, emails only update)
-7. Use canonical names in ALL output
-8. Update Aliases with newly discovered variants
-
-## Common Patterns
-
-| Pattern | Resolution |
-|---------|------------|
-| First name + same org in context | Full name at that org |
-| Email exact match | Definitive match |
-| Email domain | Resolves to organization |
-| "their CTO" + org context | Person with CTO role at org |
-| "the pilot" + org context | Project involving that org |
-| Name in Aliases field | Canonical name from that note |
-
-## Disambiguation Priority
-
-1. **Email match** — Definitive
-2. **Organization context** — Strong signal
-3. **Role match** — Good signal if org also matches
-4. **Aliases field** — Explicit match
-5. **Recency** — Weak signal, use as tiebreaker
-
-## Handling Failures
-
-| Situation | Source Type | Action |
-|-----------|-------------|--------|
-| No match + passes "Would I prep?" | Meeting | Create new note |
-| No match + passes "Would I prep?" | Email | Do NOT create (skip) |
-| No match + fails "Would I prep?" | Both | Mention in org note only |
-| Multiple matches + can disambiguate | Both | Use disambiguation rules |
-| Multiple matches + cannot disambiguate | Meeting | Create note with "(possibly same as [[X]])" |
-| Multiple matches + cannot disambiguate | Email | Skip, don't update either |
-| Conflicting information | Both | Note both versions, flag for review |
-
----
-
-# Examples
-
-## Example 1: Meeting — Creates Notes
-
-**source_file:** \`2025-01-15-meeting.md\`
-\`\`\`
-Meeting: Acme Integration Kickoff
-Date: 2025-01-15
-Attendees: Sarah Chen (sarah@acme.com), David Kim (david@acme.com), Arj (arj@rowboat.com)
-
-Transcript:
-Sarah: Thanks for meeting. We're excited about the pilot.
-David: From a technical side, we need API access first.
-Sarah: Our CTO Jennifer wants to join the next call.
-...
-\`\`\`
-
-### Step 0: Determine Source Type
-
-Has \`Meeting:\` and \`Attendees:\` → \`source_type = "meeting"\` → Can create notes
-
-### Step 1: Filter
-
-Not mass email, not automated. Continue.
-
-### Step 2: Parse
-
-- Date: 2025-01-15
-- Attendees: Sarah Chen, David Kim, Arj (self — exclude)
-- Variants: "Sarah Chen", "sarah@acme.com", "David Kim", "David", "Jennifer", "CTO", "Acme", "the pilot"
-
-### Step 3: Search Existing Notes
-\`\`\`
-workspace-grep({ pattern: "Sarah Chen", searchPath: "knowledge" })
-# Output: (none)
-
-workspace-grep({ pattern: "acme", searchPath: "knowledge" })
-# Output: (none)
-\`\`\`
-
-No existing notes. This is a new relationship.
-
-### Step 4: Resolve Entities
-
-**Resolution Map:**
-\`\`\`
-NEW ENTITIES (meeting — create):
-- "Sarah Chen" → Create [[People/Sarah Chen]]
-- "David Kim" → Create [[People/David Kim]]
-- "Jennifer" (CTO) → Create [[People/Jennifer]]
-- "Acme" → Create [[Organizations/Acme Corp]]
-- "the pilot" → Create [[Projects/Acme Integration]]
-\`\`\`
-
-### Step 5: Identify New Entities
-
-All attendees are external and pass "Would I prep?" test:
-- Sarah Chen (key contact) → Create
-- David Kim (technical contact) → Create
-- Jennifer (CTO, mentioned) → Create
-- Acme Corp (prospect company) → Create
-- Acme Integration (project) → Create
-
-### Step 6: Extract Content
-
-- Decisions: None yet
-- Commitments: Provide API access, schedule call with Jennifer
-- Key facts: Excited about pilot, need API access first, CTO involved
-
-### Step 7: Detect State Changes
-
-No existing notes → No state changes to detect.
-
-### Steps 8-10: Check, Write, Link
-
-Create all notes with extracted content, ensure bidirectional links.
-
-**Example output for Sarah Chen:**
-\`\`\`markdown
-# Sarah Chen
-
-## Info
-**Role:** Engineering (led technical discussion in kickoff meeting)
-**Organization:** [[Organizations/Acme Corp]]
-**Email:** sarah@acme.com
-**Aliases:** Sarah, sarah@acme.com
-**First met:** 2025-01-15
-**Last seen:** 2025-01-15
-
-## Summary
-Key contact at [[Organizations/Acme Corp]] for the [[Projects/Acme Integration]] pilot.
-Leading the technical evaluation. Reports to [[People/Jennifer]] (CTO).
-
-## Connected to
-- [[Organizations/Acme Corp]] — works at
-- [[People/David Kim]] — colleague
-- [[People/Jennifer]] — reports to (CTO)
-- [[Projects/Acme Integration]] — key contact
-
-## Activity
-- **2025-01-15** (meeting): Kickoff meeting for [[Projects/Acme Integration]]. Excited about pilot. [[People/David Kim]] needs API access first. [[People/Jennifer]] (CTO) joining next call.
-
-## Key facts
-- Leading technical evaluation for pilot
-- Needs API access to proceed
-- CTO Jennifer involved in next steps
-
-## Open items
-- [ ] Provide API access to [[People/David Kim]]
-- [ ] Schedule follow-up call with [[People/Jennifer]]
-\`\`\`
-
-**Example output for Acme Integration:**
-\`\`\`markdown
-# Acme Integration
-
-## Info
-**Type:** deal
-**Status:** planning
-**Started:** 2025-01-15
-**Last activity:** 2025-01-15
-
-## Summary
-Pilot integration project with [[Organizations/Acme Corp]].
-Technical evaluation phase, working toward Q2 launch.
-
-## People
-- [[People/Sarah Chen]] — key contact
-- [[People/David Kim]] — technical lead
-- [[People/Jennifer]] — CTO sponsor
-
-## Organizations
-- [[Organizations/Acme Corp]] — prospect
-
-## Timeline
-**2025-01-15** (meeting)
-Kickoff meeting. Team excited about pilot. API access needed first. CTO [[People/Jennifer]] joining next call.
-
-## Open items
-- [ ] Provide API access to [[People/David Kim]]
-- [ ] Schedule follow-up call with [[People/Jennifer]]
-\`\`\`
-
----
-
-## Example 2: Email with State Changes
-
-**source_file:** \`2025-01-20-email.md\`
-\`\`\`
-From: sarah@acme.com
-To: arj@rowboat.com
-Date: 2025-01-20
-Subject: Great news!
-
-Hi Arj,
-
-Great news — leadership approved the pilot! Legal is drafting the 
-contract now. We should be ready to kick off by end of month.
-
-Here's the API documentation you requested.
-
-Also, I've been promoted to VP of Engineering as of this month!
-
-Best,
-Sarah Chen
-VP Engineering, Acme Corp
-\`\`\`
-
-### Step 0: Determine Source Type
-
-\`source_type = "email"\` → Can only update existing notes
-
-### Step 1: Filter
-
-Check for existing relationship:
-\`\`\`
-workspace-grep({ pattern: "sarah@acme.com", searchPath: "knowledge" })
-# Output: notes/People/Sarah Chen.md
-\`\`\`
-
-Existing note found. Continue.
-
-### Steps 2-5: Parse, Search, Resolve, Skip
-
-**Resolution Map:**
-\`\`\`
-RESOLVED:
-- "Sarah", "sarah@acme.com" → [[People/Sarah Chen]]
-- "Acme" → [[Organizations/Acme Corp]]
-\`\`\`
-
-### Step 6: Extract Content
-
-- Decision: Leadership approved pilot
-- Commitment: Contract being drafted, kickoff by end of month
-- Key fact: Legal involved, targeting end of month kickoff
-
-### Step 7: Detect State Changes
-
-**7a: Project Status:**
-- "leadership approved the pilot" → Status: planning → active ✓
-
-**7b: Open Item Resolution:**
-- "Here's the API documentation you requested"
-- Existing open item: \`- [ ] Send API documentation — by Friday\`
-- Match found → Mark complete ✓
-
-**7c: Role Change:**
-- Signature: "VP Engineering"
-- Existing: "Engineering" (inferred)
-- Change detected → Update role ✓
-
-**7d: Relationship Change:**
-- "Legal is drafting the contract" → Still prospect (not signed yet)
-- No change
-
-**State Change List:**
-\`\`\`
-STATE CHANGES:
-- [[Projects/Acme Integration]]: Status planning → active
-- [[People/Sarah Chen]]: Role "Engineering" → "VP Engineering"
-- [[People/Sarah Chen]]: Open item "Provide API access" → completed (they sent docs)
-\`\`\`
-
-### Steps 8-10: Check, Write, Link
-
-**Update Sarah Chen.md:**
-\`\`\`markdown
-# Sarah Chen
-
-## Info
-**Role:** VP Engineering
-**Organization:** [[Organizations/Acme Corp]]
-**Email:** sarah@acme.com
-**Aliases:** Sarah, sarah@acme.com
-**First met:** 2025-01-15
-**Last seen:** 2025-01-20
-
-## Summary
-VP Engineering at [[Organizations/Acme Corp]] leading the [[Projects/Acme Integration]] pilot.
-Key technical decision-maker. Recently promoted.
-
-## Connected to
-- [[Organizations/Acme Corp]] — works at
-- [[People/David Kim]] — colleague
-- [[People/Jennifer]] — reports to (CTO)
-- [[Projects/Acme Integration]] — key contact
-
-## Activity
-- **2025-01-20** (email): Leadership approved pilot. [Status → active] Legal drafting contract. Kickoff by end of month. Sent API documentation. [Role → VP Engineering]
-- **2025-01-15** (meeting): Kickoff meeting for [[Projects/Acme Integration]]. Excited about pilot. [[People/David Kim]] needs API access first. [[People/Jennifer]] (CTO) joining next call.
-
-## Key facts
-- Leading technical evaluation for pilot
-- Promoted to VP Engineering (Jan 2025)
-- Legal drafting contract
-
-## Open items
-- [x] Provide API access to [[People/David Kim]] — completed 2025-01-20
-- [ ] Schedule follow-up call with [[People/Jennifer]]
-\`\`\`
-
-**Update Acme Integration.md:**
-\`\`\`markdown
-# Acme Integration
-
-## Info
-**Type:** deal
-**Status:** active
-**Started:** 2025-01-15
-**Last activity:** 2025-01-20
-
-## Summary
-Pilot integration project with [[Organizations/Acme Corp]].
-Leadership approved, contract in progress. Targeting end of month kickoff.
-
-## Timeline
-**2025-01-20** (email)
-Leadership approved pilot. [Status → active] Legal drafting contract. Targeting end of month kickoff.
-
-**2025-01-15** (meeting)
-Kickoff meeting. Team excited about pilot. API access needed first. CTO [[People/Jennifer]] joining next call.
-\`\`\`
-
----
-
-## Example 3: Email — No Existing Relationship, Skip
-
-**source_file:** \`2025-01-16-email.md\`
-\`\`\`
-From: sales@randomvendor.com
-To: arj@rowboat.com
-Date: 2025-01-16
-Subject: Quick question about your data needs
-
-Hi,
-
-I noticed your company is growing fast. Would love to show you 
-how we can help with your data infrastructure...
-
-Best,
-John Smith
-\`\`\`
-
-### Step 0: Determine Source Type
-
-\`source_type = "email"\`
-
-### Step 1: Filter
-
-Check for existing relationship:
-\`\`\`
-workspace-grep({ pattern: "randomvendor", searchPath: "knowledge" })
-# Output: (none)
-
-workspace-grep({ pattern: "John Smith", searchPath: "knowledge" })
-# Output: (none)
-\`\`\`
-
-No existing note. This is an email. Cannot create notes.
-
-**Output:**
-\`\`\`
-SKIP
-Reason: No existing relationship (email from unknown contact)
-\`\`\`
-
----
-
-## Example 4: Email — Warm Intro (Exception)
-
-**source_file:** \`2025-01-16-email.md\`
-\`\`\`
-From: david@friendly.vc
-To: arj@rowboat.com
-Cc: jennifer@newco.com
-Date: 2025-01-16
-Subject: Intro: Jennifer Lee <> Arj
-
-Hey Arj,
-
-Want to introduce you to Jennifer Lee, CEO of NewCo. She's building 
-something interesting in your space and would love to chat.
-
-Jennifer — Arj is the founder of Rowboat, doing great work on AI agents.
-
-I'll let you two take it from here!
-
-David
-\`\`\`
-
-### Step 0: Determine Source Type
-
-\`source_type = "email"\`
-
-### Step 1: Filter
-
-Check for sender:
-\`\`\`
-workspace-grep({ pattern: "david@friendly.vc", searchPath: "knowledge" })
-# Output: notes/People/David Park.md
-\`\`\`
-
-Sender exists in memory. Check if this is a warm intro:
-- Subject contains "Intro:" ✓
-- Body contains "introduce you to" ✓
-- New person (Jennifer Lee) is CC'd ✓
-
-**This is a warm intro. Exception applies.**
-
-### Steps 2-4: Parse, Search, Resolve
-
-**Resolution Map:**
-\`\`\`
-RESOLVED:
-- "David" → [[People/David Park]] (sender, exists)
-
-NEW ENTITIES (warm intro exception — create):
-- "Jennifer Lee" → Create [[People/Jennifer Lee]]
-- "NewCo" → Create [[Organizations/NewCo]]
-\`\`\`
-
-### Step 5: Create Notes (Exception)
-
-Even though this is an email, create notes for the introduced person.
-
-### Step 7: Detect State Changes
-
-No existing notes for Jennifer Lee / NewCo → No state changes.
-
-### Output
-
-Creates 2 new notes ([[People/Jennifer Lee]], [[Organizations/NewCo]]). Updates [[People/David Park]] with activity.
-
----
-
-## Example 5: Meeting — Transactional, Minimal Notes
-
-**source_file:** \`2025-01-15-meeting.md\`
-\`\`\`
-Meeting: HSBC Account Setup
-Date: 2025-01-15
-Attendees: James Wong (james@hsbc.com), Sarah Lee (sarah.lee@hsbc.com), Arj
-
-Transcript:
-James: Let's go through the account setup process.
-Sarah: I'll handle the wire transfer limits after.
-...
-\`\`\`
-
-### Step 0: Determine Source Type
-
-\`source_type = "meeting"\` → Can create notes
-
-### Step 5: Identify New Entities
-
-Apply "Would I prep?" test:
-- James Wong (bank RM) → No
-- Sarah Lee (support) → No
-- HSBC (organization) → Yes, worth one org note
-
-**Action:** Create org note only, list people in Contacts section.
-
-### Output
-\`\`\`markdown
-# HSBC
-
-## Info
-**Type:** company
-**Industry:** Banking
-**Relationship:** vendor (banking)
-**Domain:** hsbc.com
-**Aliases:** HSBC Bank
-**First met:** 2025-01-15
-**Last seen:** 2025-01-15
-
-## Summary
-Business banking provider. Account setup completed January 2025.
-
-## People
-
-## Contacts
-- James Wong — Relationship Manager, account setup
-- Sarah Lee — Support, wire transfer limits
-
-## Activity
-- **2025-01-15** (meeting): Account setup walkthrough. Wire transfer limits discussed.
-
-## Key facts
-- Account Number: XXXX-1234
-- Daily wire limit: $50,000
-
-## Open items
-\`\`\`
-
----
-
-# Summary: The Core Rules
-
-| Source Type | Creates Notes? | Updates Notes? | Detects State Changes? |
-|-------------|---------------|----------------|------------------------|
-| Meeting | Yes | Yes | Yes |
-| Voice memo | Yes | Yes | Yes |
-| Email (known contact) | No | Yes | Yes |
-| Email (unknown contact) | No | No (SKIP) | No |
-| Email (warm intro) | Yes (exception) | Yes | Yes |
-
-**Voice memo activity format:** Always include a link to the source voice memo:
-\`\`\`
-**2025-01-15** (voice memo): Discussed project timeline with [[People/Sarah Chen]]. See [[Voice Memos/2025-01-15/voice-memo-...]]
-\`\`\`
-
----
-
-# State Change Reference
-
-## What Changes Automatically
-
-| Field | Trigger | Example |
-|-------|---------|---------|
-| Project Status | "approved", "on hold", "cancelled", "completed" | planning → active |
-| Open Items | "here's the X you requested", "sent the X" | [ ] → [x] |
-| Person Role | New title in signature, "promoted to" | Engineer → VP |
-| Org Relationship | "signed contract", "now a customer" | prospect → customer |
-| Person Organization | "I've joined X", new email domain | Acme → NewCo |
-
-## How to Log State Changes
-
-In activity entries, use \`[Field → value]\` notation:
-\`\`\`markdown
-- **2025-01-20** (email): Leadership approved. [Status → active] Contract in progress.
-\`\`\`
-
-## When NOT to Change State
-
-- Signal is ambiguous ("might move forward")
-- Contradicts recent information (check activity log)
-- Would be a regression (active → planning)
-- Based on speculation, not explicit statement
-
----
-
-# Error Handling
-
-1. **Missing data:** Leave blank rather than writing "Unknown"
-2. **Ambiguous names:** For meetings, create note with "(possibly same as [[X]])". For emails, skip.
-3. **Conflicting info:** Note both versions, mark "needs clarification"
-4. **grep returns nothing:** For meetings, apply qualifying rules and create if appropriate. For emails, skip.
-5. **State change unclear:** Log in activity but don't change the field
-6. **Note file malformed:** Log warning, attempt partial update, continue
-7. **Shell command fails:** Log error, continue with what you have
-
----
-
-# Quality Checklist
-
-Before completing, verify:
-
-**Source Type:**
-- [ ] Correctly identified as meeting or email
-- [ ] Applied correct rules (meetings create, emails only update)
-
-**Resolution:**
-- [ ] Extracted all name variants from source
-- [ ] Searched notes including Aliases fields
-- [ ] Built resolution map before writing
-- [ ] Used absolute paths \`[[Folder/Name]]\` in ALL links
-- [ ] Updated Aliases fields with new variants discovered
-
-**Filtering:**
-- [ ] Excluded self (user.name, user.email, @user.domain)
-- [ ] Applied "Would I prep?" test to each person
-- [ ] Transactional contacts in Org Contacts, not People notes
-- [ ] Source correctly classified (process vs skip)
-- [ ] Emails from unknown contacts skipped (unless warm intro)
-
-**Content Quality:**
-- [ ] Summaries describe relationship, not communication method
-- [ ] Roles inferred where possible (with qualifier)
-- [ ] Key facts are substantive (no "name only provided" filler)
-- [ ] Open items are commitments/next steps (no "find their email" tasks)
-- [ ] Empty sections left empty rather than filled with placeholders
-
-**State Changes:**
-- [ ] Detected project status changes
-- [ ] Marked completed open items with [x]
-- [ ] Updated roles if changed
-- [ ] Updated relationships if changed
-- [ ] Logged all state changes in activity with [Field → value] notation
-- [ ] Only applied clear, unambiguous state changes
-
-**Structure:**
-- [ ] All entity mentions use \`[[Folder/Name]]\` absolute links
-- [ ] Activity entries are reverse chronological
-- [ ] No duplicate activity entries
-- [ ] Dates are YYYY-MM-DD
-- [ ] Bidirectional links are consistent
-- [ ] New notes in correct folders
-`;
\ No newline at end of file
diff --git a/apps/x/packages/core/src/knowledge/note_creation_low.ts b/apps/x/packages/core/src/knowledge/note_creation_low.ts
deleted file mode 100644
index 29922fce..00000000
--- a/apps/x/packages/core/src/knowledge/note_creation_low.ts
+++ /dev/null
@@ -1,874 +0,0 @@
-export const raw = `---
-model: gpt-5.2
-tools:
-  workspace-writeFile:
-    type: builtin
-    name: workspace-writeFile
-  workspace-readFile:
-    type: builtin
-    name: workspace-readFile
-  workspace-edit:
-    type: builtin
-    name: workspace-edit
-  workspace-readdir:
-    type: builtin
-    name: workspace-readdir
-  workspace-mkdir:
-    type: builtin
-    name: workspace-mkdir
-  workspace-grep:
-    type: builtin
-    name: workspace-grep
-  workspace-glob:
-    type: builtin
-    name: workspace-glob
----
-# Task
-
-You are a memory agent. Given a single source file (email, meeting transcript, or voice memo), you will:
-
-1. **Determine source type (meeting or email)**
-2. **Evaluate if the source is worth processing**
-3. **Search for all existing related notes**
-4. **Resolve entities to canonical names**
-5. Identify new entities worth tracking
-6. Extract structured information (decisions, commitments, key facts)
-7. **Detect state changes (status updates, resolved items, role changes)**
-8. Create new notes or update existing notes
-9. **Apply state changes to existing notes**
-
-The core rule: **Capture broadly. Meetings, voice memos, and emails create notes for most external contacts.**
-
-You have full read access to the existing knowledge directory. Use this extensively to:
-- Find existing notes for people, organizations, projects mentioned
-- Resolve ambiguous names (find existing note for "David")
-- Understand existing relationships before updating
-- Avoid creating duplicate notes
-- Maintain consistency with existing content
-- **Detect when new information changes the state of existing notes**
-
-# Inputs
-
-1. **source_file**: Path to a single file to process (email or meeting transcript)
-2. **knowledge_folder**: Path to Obsidian vault (read/write access)
-3. **user**: Information about the owner of this memory
-   - name: e.g., "Arj"
-   - email: e.g., "arj@rowboat.com"
-   - domain: e.g., "rowboat.com"
-4. **knowledge_index**: A pre-built index of all existing notes (provided in the message)
-
-# Knowledge Base Index
-
-**IMPORTANT:** You will receive a pre-built index of all existing notes at the start of each request. This index contains:
-- All people notes with their names, emails, aliases, and organizations
-- All organization notes with their names, domains, and aliases
-- All project notes with their names and statuses
-- All topic notes with their names and keywords
-
-**USE THE INDEX for entity resolution instead of grep/search commands.** This is much faster.
-
-When you need to:
-- Check if a person exists → Look up by name/email/alias in the index
-- Find an organization → Look up by name/domain in the index
-- Resolve "David" to a full name → Check index for people with that name/alias + organization context
-
-**Only use \`cat\` to read full note content** when you need details not in the index (e.g., existing activity logs, open items).
-
-# Tools Available
-
-You have access to these tools:
-
-**For reading files:**
-\`\`\`
-workspace-readFile({ path: "knowledge/People/Sarah Chen.md" })
-\`\`\`
-
-**For creating NEW files:**
-\`\`\`
-workspace-writeFile({ path: "knowledge/People/Sarah Chen.md", data: "# Sarah Chen\\n\\n..." })
-\`\`\`
-
-**For editing EXISTING files (preferred for updates):**
-\`\`\`
-workspace-edit({
-  path: "knowledge/People/Sarah Chen.md",
-  oldString: "## Activity\\n",
-  newString: "## Activity\\n- **2026-02-03** (meeting): New activity entry\\n"
-})
-\`\`\`
-
-**For listing directories:**
-\`\`\`
-workspace-readdir({ path: "knowledge/People" })
-\`\`\`
-
-**For creating directories:**
-\`\`\`
-workspace-mkdir({ path: "knowledge/Projects", recursive: true })
-\`\`\`
-
-**For searching files:**
-\`\`\`
-workspace-grep({ pattern: "Acme Corp", searchPath: "knowledge", fileGlob: "*.md" })
-\`\`\`
-
-**For finding files by pattern:**
-\`\`\`
-workspace-glob({ pattern: "**/*.md", cwd: "knowledge/People" })
-\`\`\`
-
-**IMPORTANT:**
-- Use \`workspace-edit\` for updating existing notes (adding activity, updating fields)
-- Use \`workspace-writeFile\` only for creating new notes
-- Prefer the knowledge_index for entity resolution (it's faster than grep)
-
-# Output
-
-Either:
-- **SKIP** with reason, if source should be ignored
-- Updated or new markdown files in notes_folder
-
----
-
-# The Core Rule: Low Strictness - Capture Broadly
-
-**LOW STRICTNESS MODE**
-
-This mode prioritizes comprehensive capture over selectivity. The goal is to never miss a potentially important contact.
-
-**Meetings create notes for:**
-- All external attendees (anyone not @user.domain)
-
-**Emails create notes for:**
-- Any personalized email from an identifiable sender
-- Anyone who reaches out directly
-- Any external contact who communicates with you
-
-**Only skip:**
-- Obvious automated/system emails (no human sender)
-- Mass newsletters with unsubscribe links
-- Truly anonymous or unidentifiable senders
-
-**Philosophy:** It's better to have a note you don't need than to miss tracking someone important.
-
----
-
-# Step 0: Determine Source Type
-
-Read the source file and determine if it's a meeting or email.
-\`\`\`
-workspace-readFile({ path: "{source_file}" })
-\`\`\`
-
-**Meeting indicators:**
-- Has \`Attendees:\` field
-- Has \`Meeting:\` title
-- Transcript format with speaker labels
-
-**Email indicators:**
-- Has \`From:\` and \`To:\` fields
-- Has \`Subject:\` field
-- Email signature
-
-**Voice memo indicators:**
-- Has \`**Type:** voice memo\` field
-- Has \`**Path:**\` field with path like \`Voice Memos/YYYY-MM-DD/...\`
-- Has \`## Transcript\` section
-
-**Set processing mode:**
-- \`source_type = "meeting"\` → Create notes for all external attendees
-- \`source_type = "email"\` → Create notes for sender if identifiable human
-- \`source_type = "voice_memo"\` → Create notes for all mentioned entities (treat like meetings)
-
----
-
-## Calendar Invite Emails
-
-Emails containing calendar invites (\`.ics\` attachments) are **high signal** - a scheduled meeting means this person matters.
-
-**How to identify:**
-- Subject contains "Invitation:", "Accepted:", "Declined:", or "Updated:"
-- Has \`.ics\` attachment reference
-
-**Rules:**
-1. **CREATE a note for the primary contact** - the person you're meeting with
-2. **Skip automated notifications** - from calendar-no-reply@google.com with no human sender
-3. **Skip "Accepted/Declined" responses** - just RSVP confirmations
-
-Once a note exists, subsequent emails will enrich it. When the meeting happens, the transcript adds more detail.
-
----
-
-# Step 1: Source Filtering (Minimal)
-
-## Skip Only These Sources
-
-### Mass Newsletters
-
-**Indicators (must have MULTIPLE of these):**
-- Unsubscribe link in body or footer
-- From a marketing address (noreply@, newsletter@, marketing@)
-- Sent to multiple recipients or undisclosed-recipients
-- Sent via marketing platforms (via sendgrid, via mailchimp, etc.)
-
-**Action:** SKIP with reason "Mass newsletter"
-
-### Purely Automated (No Human Sender)
-
-**Indicators:**
-- From automated systems with no human behind them (alerts@, notifications@)
-- Password resets, login alerts
-- System notifications (GitHub automated, CI/CD alerts)
-- Receipt confirmations with no human contact info
-
-**Action:** SKIP with reason "Automated system message"
-
-### Truly Low-Signal
-
-**Indicators (must be clearly content-free):**
-- Body is ONLY "Thanks!", "Got it", "OK" with nothing else
-- Auto-replies ("I'm out of office") with no human context
-
-**Action:** SKIP with reason "No substantive content"
-
-## Process Everything Else
-
-**Important:** When in doubt, PROCESS. In low strictness mode, we err on the side of capturing more.
-
-If skipping:
-\`\`\`
-SKIP
-Reason: {reason}
-\`\`\`
-
-If processing, continue to Step 2.
-
----
-
-# Step 2: Read and Parse Source File
-\`\`\`
-workspace-readFile({ path: "{source_file}" })
-\`\`\`
-
-Extract metadata:
-
-**For meetings:**
-- **Date:** From header or filename
-- **Title:** Meeting name
-- **Attendees:** List of participants
-- **Duration:** If available
-
-**For emails:**
-- **Date:** From \`Date:\` header
-- **Subject:** From \`Subject:\` header
-- **From:** Sender email/name
-- **To/Cc:** Recipients
-
-## 2a: Exclude Self
-
-Never create or update notes for:
-- The user (matches user.name, user.email, or @user.domain)
-- Anyone @{user.domain} (colleagues at user's company)
-
-Filter these out from attendees/participants before proceeding.
-
-## 2b: Extract All Name Variants
-
-From the source, collect every way entities are referenced:
-
-**People variants:**
-- Full names: "Sarah Chen"
-- First names only: "Sarah"
-- Last names only: "Chen"
-- Initials: "S. Chen"
-- Email addresses: "sarah@acme.com"
-- Roles/titles: "their CTO", "the VP of Engineering"
-
-**Organization variants:**
-- Full names: "Acme Corporation"
-- Short names: "Acme"
-- Abbreviations: "AC"
-- Email domains: "@acme.com"
-
-**Project variants:**
-- Explicit names: "Project Atlas"
-- Descriptive references: "the integration", "the pilot", "the deal"
-
-Create a list of all variants found.
-
----
-
-# Step 3: Look Up Existing Notes in Index
-
-**Use the provided knowledge_index to find existing notes. Do NOT use grep commands.**
-
-## 3a: Look Up People
-
-For each person variant (name, email, alias), check the index:
-
-\`\`\`
-From index, find matches for:
-- "Sarah Chen" → Check People table for matching name
-- "Sarah" → Check People table for matching name or alias
-- "sarah@acme.com" → Check People table for matching email
-- "@acme.com" → Check People table for matching organization or check Organizations for domain
-\`\`\`
-
-## 3b: Look Up Organizations
-
-\`\`\`
-From index, find matches for:
-- "Acme Corp" → Check Organizations table for matching name
-- "Acme" → Check Organizations table for matching name or alias
-- "acme.com" → Check Organizations table for matching domain
-\`\`\`
-
-## 3c: Look Up Projects and Topics
-
-\`\`\`
-From index, find matches for:
-- "the pilot" → Check Projects table for related names
-- "SOC 2" → Check Topics table for matching keywords
-\`\`\`
-
-## 3d: Read Full Notes When Needed
-
-Only read the full note content when you need details not in the index (e.g., activity logs, open items):
-\`\`\`
-workspace-readFile({ path: "{knowledge_folder}/People/Sarah Chen.md" })
-\`\`\`
-
-**Why read these notes:**
-- Find canonical names (David → David Kim)
-- Check Aliases fields for known variants
-- Understand existing relationships
-- See organization context for disambiguation
-- Check what's already captured (avoid duplicates)
-- Review open items (some might be resolved)
-- **Check current status fields (might need updating)**
-- **Check current roles (might have changed)**
-
-## 3e: Matching Criteria
-
-Use these criteria to determine if a variant matches an existing note:
-
-**People matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| First name "Sarah" | Full name "Sarah Chen" | Same organization context |
-| Email "sarah@acme.com" | Email field | Exact match |
-| Email domain "@acme.com" | Organization "Acme Corp" | Domain matches org |
-| Role "VP Engineering" | Role field | Same org + same role |
-| First name + company context | Full name + Organization | Company matches |
-| Any variant | Aliases field | Listed in aliases |
-
-**Organization matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| "Acme" | "Acme Corp" | Substring match |
-| "Acme Corporation" | "Acme Corp" | Same root name |
-| "@acme.com" | Domain field | Domain matches |
-| Any variant | Aliases field | Listed in aliases |
-
-**Project matching:**
-
-| Source has | Note has | Match if |
-|------------|----------|----------|
-| "the pilot" | "Acme Pilot" | Same org context in source |
-| "integration project" | "Acme Integration" | Same org + similar type |
-| "Series A" | "Series A Fundraise" | Unique identifier match |
-
----
-
-# Step 4: Resolve Entities to Canonical Names
-
-Using the search results from Step 3, resolve each variant to a canonical name.
-
-## 4a: Build Resolution Map
-
-Create a mapping from every source reference to its canonical form.
-
-## 4b: Apply Source Type Rules (Low Strictness)
-
-**If source_type == "meeting":**
-- Resolved entities → Update existing notes
-- New entities → Create new notes for ALL external attendees
-
-**If source_type == "email" (LOW STRICTNESS):**
-- Resolved entities → Update existing notes
-- New entities → Create notes for the sender and any mentioned contacts
-
-## 4c: Disambiguation Rules
-
-When multiple candidates match a variant, disambiguate by:
-1. Email match (definitive)
-2. Organization context (strong signal)
-3. Role match
-4. Recency (tiebreaker)
-
-## 4d: Resolution Map Output
-
-Final resolution map before proceeding:
-\`\`\`
-RESOLVED (use canonical name with absolute path):
-- "Sarah", "Sarah Chen", "sarah@acme.com" → [[People/Sarah Chen]]
-
-NEW ENTITIES (create notes):
-- "Jennifer" (CTO, Acme Corp) → Create [[People/Jennifer]]
-
-AMBIGUOUS (create with disambiguation note):
-- "Mike" (no context) → Create [[People/Mike]] with note about ambiguity
-\`\`\`
-
----
-
-# Step 5: Identify New Entities (Low Strictness - Capture Broadly)
-
-For entities not resolved to existing notes, create notes for most of them.
-
-## People
-
-### Who Gets a Note (Low Strictness)
-
-**CREATE a note for:**
-- ALL external meeting attendees (not @user.domain)
-- ALL email senders with identifiable names/emails
-- Anyone CC'd on emails who seems relevant
-- Anyone mentioned by name in conversations
-- Cold outreach senders (even if unsolicited)
-- Sales reps, recruiters, service providers
-- Anyone who might be useful to remember later
-
-**DO NOT create notes for:**
-- Internal colleagues (@user.domain)
-- Truly anonymous/unidentifiable senders
-- System-generated sender names with no human behind them
-
-### The Low Strictness Test
-
-Ask: Could this person ever be useful to remember?
-
-- Sarah Chen, VP Engineering → **Yes, create note**
-- James from HSBC → **Yes, create note** (might need banking help again)
-- Random recruiter → **Yes, create note** (might want to contact later)
-- Cold sales person → **Yes, create note** (might be relevant someday)
-- Support rep → **Yes, create note** (might need them again)
-
-### Role Inference
-
-If role is not explicitly stated, infer from context. Write "Unknown" only if truly impossible to infer anything.
-
-### Relationship Type Guide (Low Strictness)
-
-| Relationship Type | Create People Notes? | Create Org Note? |
-|-------------------|----------------------|------------------|
-| Customer | Yes — all contacts | Yes |
-| Prospect | Yes — all contacts | Yes |
-| Investor | Yes | Yes |
-| Partner | Yes — all contacts | Yes |
-| Vendor | Yes — all contacts | Yes |
-| Bank/Financial | Yes | Yes |
-| Candidate | Yes | No |
-| Recruiter | Yes | Optional |
-| Service provider | Yes | Optional |
-| Cold outreach | Yes | Optional |
-| Support interaction | Yes | Optional |
-
-## Organizations
-
-**CREATE a note if:**
-- Anyone from that org is mentioned or contacted you
-- The org is mentioned in any context
-
-**Only skip:**
-- Organizations you genuinely can't identify
-
-## Projects
-
-**CREATE a note if:**
-- Discussed in meeting or email
-- Any indication of ongoing work or collaboration
-
-## Topics
-
-**CREATE a note if:**
-- Mentioned more than once
-- Seems like a recurring theme
-
----
-
-# Step 6: Extract Content
-
-For each entity that has or will have a note, extract relevant content.
-
-## Decisions
-
-Extract what was decided, when, by whom, and why.
-
-## Commitments
-
-Extract who committed to what, and any deadlines.
-
-## Key Facts
-
-Key facts should be **substantive information** — not commentary about missing data.
-
-**Extract if:**
-- Specific numbers, dates, or metrics
-- Preferences or working style
-- Background information
-- Authority or decision process
-- Concerns or constraints
-- What they're working on or interested in
-
-**Never include:**
-- Meta-commentary about missing data
-- Obvious facts already in Info section
-- Placeholder text
-
-**If there are no substantive key facts, leave the section empty.**
-
-## Open Items
-
-**Include:**
-- Commitments made
-- Requests received
-- Next steps discussed
-- Follow-ups agreed
-
-**Never include:**
-- Data gaps or research tasks
-- Wishes or hypotheticals
-
-## Summary
-
-The summary should answer: **"Who is this person and why do I know them?"**
-
-Write 2-3 sentences covering their role/function, context of the relationship, and what you're discussing.
-
-## Activity Summary
-
-One line summarizing this source's relevance to the entity:
-\`\`\`
-**{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[links]]}
-\`\`\`
-
-**For voice memos:** Include a link to the voice memo file using the Path field:
-\`\`\`
-**2025-01-15** (voice memo): Discussed [[Projects/Acme Integration]] timeline. See [[Voice Memos/2025-01-15/voice-memo-2025-01-15T10-30-00-000Z]]
-\`\`\`
-
----
-
-# Step 7: Detect State Changes
-
-Review the extracted content for signals that existing note fields should be updated.
-
-## 7a: Project Status Changes
-
-Look for signals like "approved", "on hold", "cancelled", "completed", etc.
-
-## 7b: Open Item Resolution
-
-Look for signals that tracked items are now complete.
-
-## 7c: Role/Title Changes
-
-Look for new titles in signatures or explicit announcements.
-
-## 7d: Organization/Relationship Changes
-
-Look for company changes, partnership announcements, etc.
-
-## 7e: Build State Change List
-
-Compile all detected state changes before writing.
-
----
-
-# Step 8: Check for Duplicates and Conflicts
-
-Before writing:
-- Check if already processed this source
-- Skip duplicate key facts
-- Handle conflicting information by noting both versions
-
----
-
-# Step 9: Write Updates
-
-## 9a: Create and Update Notes
-
-**IMPORTANT: Write sequentially, one file at a time.**
-- Generate content for exactly one note.
-- Issue exactly one write/edit command.
-- Wait for the tool to return before generating the next note.
-- Do NOT batch multiple write commands in a single response.
-
-**For NEW entities (use workspace-writeFile):**
-\`\`\`
-workspace-writeFile({
-  path: "{knowledge_folder}/People/Jennifer.md",
-  data: "# Jennifer\\n\\n## Summary\\n..."
-})
-\`\`\`
-
-**For EXISTING entities (use workspace-edit):**
-- Read current content first with workspace-readFile
-- Use workspace-edit to add activity entry at TOP (reverse chronological)
-- Update fields using targeted edits
-\`\`\`
-workspace-edit({
-  path: "{knowledge_folder}/People/Sarah Chen.md",
-  oldString: "## Activity\\n",
-  newString: "## Activity\\n- **2026-02-03** (meeting): Met to discuss project timeline\\n"
-})
-\`\`\`
-
-## 9b: Apply State Changes
-
-Update all fields identified in Step 7.
-
-## 9c: Update Aliases
-
-Add newly discovered name variants to Aliases field.
-
-## 9d: Writing Rules
-
-- **Always use absolute paths** with format \`[[Folder/Name]]\` for all links
-- Use YYYY-MM-DD format for dates
-- Be concise: one line per activity entry
-- Escape quotes properly in shell commands
-- Write only one file per response (no multi-file write batches)
-
----
-
-# Step 10: Ensure Bidirectional Links
-
-After writing, verify links go both ways.
-
-## Absolute Link Format
-
-**IMPORTANT:** Always use absolute links:
-\`\`\`markdown
-[[People/Sarah Chen]]
-[[Organizations/Acme Corp]]
-[[Projects/Acme Integration]]
-[[Topics/Security Compliance]]
-\`\`\`
-
-## Bidirectional Link Rules
-
-| If you add... | Then also add... |
-|---------------|------------------|
-| Person → Organization | Organization → Person |
-| Person → Project | Project → Person |
-| Project → Organization | Organization → Project |
-| Project → Topic | Topic → Project |
-| Person → Person | Person → Person (reverse) |
-
----
-
-# Note Templates
-
-## People
-\`\`\`markdown
-# {Full Name}
-
-## Info
-**Role:** {role, inferred role, or Unknown}
-**Organization:** [[Organizations/{organization}]] or leave blank
-**Email:** {email or leave blank}
-**Aliases:** {comma-separated: first name, nicknames, email}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: Who they are, why you know them.}
-
-## Connected to
-- [[Organizations/{Organization}]] — works at
-- [[People/{Person}]] — {relationship}
-- [[Projects/{Project}]] — {role}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-\`\`\`
-
-## Organizations
-\`\`\`markdown
-# {Organization Name}
-
-## Info
-**Type:** {company|team|institution|other}
-**Industry:** {industry or leave blank}
-**Relationship:** {customer|prospect|partner|competitor|vendor|other}
-**Domain:** {primary email domain}
-**Aliases:** {short names, abbreviations}
-**First met:** {YYYY-MM-DD}
-**Last seen:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this org is, what your relationship is.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Contacts
-{For contacts who have their own notes}
-
-## Projects
-- [[Projects/{Project}]] — {relationship}
-
-## Activity
-- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary}
-
-## Key facts
-{Substantive facts only. Leave empty if none.}
-
-## Open items
-{Commitments and next steps only. Leave empty if none.}
-\`\`\`
-
-## Projects
-\`\`\`markdown
-# {Project Name}
-
-## Info
-**Type:** {deal|product|initiative|hiring|other}
-**Status:** {active|planning|on hold|completed|cancelled}
-**Started:** {YYYY-MM-DD or leave blank}
-**Last activity:** {YYYY-MM-DD}
-
-## Summary
-{2-3 sentences: What this project is, goal, current state.}
-
-## People
-- [[People/{Person}]] — {role}
-
-## Organizations
-- [[Organizations/{Org}]] — {relationship}
-
-## Related
-- [[Topics/{Topic}]] — {relationship}
-
-## Timeline
-**{YYYY-MM-DD}** ({meeting|email|voice memo})
-{What happened.}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}
-
-## Open items
-{Commitments and next steps only.}
-
-## Key facts
-{Substantive facts only.}
-\`\`\`
-
-## Topics
-\`\`\`markdown
-# {Topic Name}
-
-## About
-{1-2 sentences: What this topic covers.}
-
-**Keywords:** {comma-separated}
-**Aliases:** {other references}
-**First mentioned:** {YYYY-MM-DD}
-**Last mentioned:** {YYYY-MM-DD}
-
-## Related
-- [[People/{Person}]] — {relationship}
-- [[Organizations/{Org}]] — {relationship}
-- [[Projects/{Project}]] — {relationship}
-
-## Log
-**{YYYY-MM-DD}** ({meeting|email}: {title})
-{Summary}
-
-## Decisions
-- **{YYYY-MM-DD}**: {Decision}
-
-## Open items
-{Commitments and next steps only.}
-
-## Key facts
-{Substantive facts only.}
-\`\`\`
-
----
-
-# Summary: Low Strictness Rules
-
-| Source Type | Creates Notes? | Updates Notes? | Detects State Changes? |
-|-------------|---------------|----------------|------------------------|
-| Meeting | Yes — ALL external attendees | Yes | Yes |
-| Voice memo | Yes — all mentioned entities | Yes | Yes |
-| Email (any human sender) | Yes | Yes | Yes |
-| Email (automated/newsletter) | No (SKIP) | No | No |
-
-**Voice memo activity format:** Always include a link to the source voice memo:
-\`\`\`
-**2025-01-15** (voice memo): Discussed project timeline with [[People/Sarah Chen]]. See [[Voice Memos/2025-01-15/voice-memo-...]]
-\`\`\`
-
-**Philosophy:** Capture broadly, filter later if needed.
-
----
-
-# Error Handling
-
-1. **Missing data:** Leave blank or write "Unknown"
-2. **Ambiguous names:** Create note with disambiguation note
-3. **Conflicting info:** Note both versions
-4. **grep returns nothing:** Create new notes
-5. **State change unclear:** Log in activity but don't change the field
-6. **Note file malformed:** Log warning, attempt partial update
-7. **Shell command fails:** Log error, continue
-
----
-
-# Quality Checklist
-
-Before completing, verify:
-
-**Source Type:**
-- [ ] Correctly identified as meeting or email
-- [ ] Applied low strictness rules (capture broadly)
-
-**Resolution:**
-- [ ] Extracted all name variants
-- [ ] Searched existing notes
-- [ ] Built resolution map
-- [ ] Used absolute paths \`[[Folder/Name]]\`
-
-**Filtering:**
-- [ ] Excluded only self and @user.domain
-- [ ] Created notes for all external contacts
-- [ ] Only skipped obvious automated/newsletters
-
-**Content Quality:**
-- [ ] Summaries describe relationship
-- [ ] Roles inferred where possible
-- [ ] Key facts are substantive
-- [ ] Open items are commitments/next steps
-
-**State Changes:**
-- [ ] Detected and applied state changes
-- [ ] Logged changes in activity
-
-**Structure:**
-- [ ] All links use \`[[Folder/Name]]\` format
-- [ ] Activity entries reverse chronological
-- [ ] Dates are YYYY-MM-DD
-- [ ] Bidirectional links consistent
-`;
\ No newline at end of file
diff --git a/apps/x/packages/core/src/knowledge/note_system.ts b/apps/x/packages/core/src/knowledge/note_system.ts
new file mode 100644
index 00000000..210d3501
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/note_system.ts
@@ -0,0 +1,202 @@
+import path from "path";
+import fs from "fs";
+import { WorkDir } from "../config/config.js";
+
+export interface NoteTypeDefinition {
+  type: string;
+  folder: string;
+  template: string;
+  extractionGuide: string;
+}
+
+// ── Default definitions (used to seed ~/.rowboat/config/notes.json) ──────────
+
+const DEFAULT_NOTE_TYPE_DEFINITIONS: NoteTypeDefinition[] = [
+  {
+    type: "People",
+    folder: "People",
+    template: `# {Full Name}
+
+## Info
+**Role:** {role, or inferred role with qualifier, or leave blank if truly unknown}
+**Organization:** [[Organizations/{organization}]] or leave blank
+**Email:** {email or leave blank}
+**Aliases:** {comma-separated: first name, nicknames, email}
+**First met:** {YYYY-MM-DD}
+**Last seen:** {YYYY-MM-DD}
+
+## Summary
+{2-3 sentences: Who they are, why you know them, what you're working on together.}
+
+## Connected to
+- [[Organizations/{Organization}]] — works at
+- [[People/{Person}]] — {colleague, introduced by, reports to}
+- [[Projects/{Project}]] — {role}
+
+## Activity
+- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links}
+
+## Key facts
+{Substantive facts only. Leave empty if none.}
+
+## Open items
+{Commitments and next steps only. Leave empty if none.}`,
+    extractionGuide:
+      "Look for: name, role, organization, email, aliases, relationship context",
+  },
+  {
+    type: "Organizations",
+    folder: "Organizations",
+    template: `# {Organization Name}
+
+## Info
+**Type:** {company|team|institution|other}
+**Industry:** {industry or leave blank}
+**Relationship:** {customer|prospect|partner|competitor|vendor|other}
+**Domain:** {primary email domain}
+**Aliases:** {comma-separated: short names, abbreviations}
+**First met:** {YYYY-MM-DD}
+**Last seen:** {YYYY-MM-DD}
+
+## Summary
+{2-3 sentences: What this org is, what your relationship is.}
+
+## People
+- [[People/{Person}]] — {role}
+
+## Contacts
+{For transactional contacts who don't get their own notes}
+
+## Projects
+- [[Projects/{Project}]] — {relationship}
+
+## Activity
+- **{YYYY-MM-DD}** ({meeting|email|voice memo}): {Summary with [[Folder/Name]] links}
+
+## Key facts
+{Substantive facts only. Leave empty if none.}
+
+## Open items
+{Commitments and next steps only. Leave empty if none.}`,
+    extractionGuide:
+      "Look for: organization name, type, industry, relationship, domain, key people, projects",
+  },
+  {
+    type: "Projects",
+    folder: "Projects",
+    template: `# {Project Name}
+
+## Info
+**Type:** {deal|product|initiative|hiring|other}
+**Status:** {active|planning|on hold|completed|cancelled}
+**Started:** {YYYY-MM-DD or leave blank}
+**Last activity:** {YYYY-MM-DD}
+
+## Summary
+{2-3 sentences: What this project is, goal, current state.}
+
+## People
+- [[People/{Person}]] — {role}
+
+## Organizations
+- [[Organizations/{Org}]] — {customer|partner|etc.}
+
+## Related
+- [[Topics/{Topic}]] — {relationship}
+- [[Projects/{Project}]] — {relationship}
+
+## Timeline
+**{YYYY-MM-DD}** ({meeting|email})
+{What happened.}
+
+## Decisions
+- **{YYYY-MM-DD}**: {Decision}. {Rationale}.
+
+## Open items
+{Commitments and next steps only. Leave empty if none.}
+
+## Key facts
+{Substantive facts only. Leave empty if none.}`,
+    extractionGuide:
+      "Look for: project name, type, status, people involved, organizations, timeline, decisions",
+  },
+  {
+    type: "Topics",
+    folder: "Topics",
+    template: `# {Topic Name}
+
+## About
+{1-2 sentences: What this topic covers.}
+
+**Keywords:** {comma-separated}
+**Aliases:** {other ways this topic is referenced}
+**First mentioned:** {YYYY-MM-DD}
+**Last mentioned:** {YYYY-MM-DD}
+
+## Related
+- [[People/{Person}]] — {relationship}
+- [[Organizations/{Org}]] — {relationship}
+- [[Projects/{Project}]] — {relationship}
+
+## Log
+**{YYYY-MM-DD}** ({meeting|email}: {title})
+{Summary with [[Folder/Name]] links}
+
+## Decisions
+- **{YYYY-MM-DD}**: {Decision}
+
+## Open items
+{Commitments and next steps only. Leave empty if none.}
+
+## Key facts
+{Substantive facts only. Leave empty if none.}`,
+    extractionGuide:
+      "Look for: topic name, keywords, related people/orgs/projects, decisions, key facts",
+  },
+];
+
+// ── Disk-backed config with mtime caching ──────────────────────────────────
+
+export const NOTES_CONFIG_PATH = path.join(WorkDir, "config", "notes.json");
+
+let cachedNoteTypeDefinitions: NoteTypeDefinition[] | null = null;
+let cachedMtimeMs: number | null = null;
+
+function ensureNotesConfigSync(): void {
+  if (!fs.existsSync(NOTES_CONFIG_PATH)) {
+    fs.writeFileSync(
+      NOTES_CONFIG_PATH,
+      JSON.stringify(DEFAULT_NOTE_TYPE_DEFINITIONS, null, 2) + "\n",
+      "utf8",
+    );
+  }
+}
+
+export function getNoteTypeDefinitions(): NoteTypeDefinition[] {
+  ensureNotesConfigSync();
+  try {
+    const stats = fs.statSync(NOTES_CONFIG_PATH);
+    if (cachedNoteTypeDefinitions && cachedMtimeMs === stats.mtimeMs) {
+      return cachedNoteTypeDefinitions;
+    }
+    const content = fs.readFileSync(NOTES_CONFIG_PATH, "utf8");
+    cachedNoteTypeDefinitions = JSON.parse(content);
+    cachedMtimeMs = stats.mtimeMs;
+    return cachedNoteTypeDefinitions!;
+  } catch {
+    cachedNoteTypeDefinitions = null;
+    cachedMtimeMs = null;
+    return DEFAULT_NOTE_TYPE_DEFINITIONS;
+  }
+}
+
+// ── Render helper ────────────────────────────────────────────────────────
+
+export function renderNoteTypesBlock(): string {
+  const defs = getNoteTypeDefinitions();
+  const sections = defs.map(
+    (d) =>
+      `## ${d.type}\n\`\`\`markdown\n${d.template}\n\`\`\``,
+  );
+  return `# Note Templates\n\n${sections.join("\n\n")}`;
+}
diff --git a/apps/x/packages/core/src/knowledge/note_tagging_agent.ts b/apps/x/packages/core/src/knowledge/note_tagging_agent.ts
new file mode 100644
index 00000000..94cd5016
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/note_tagging_agent.ts
@@ -0,0 +1,132 @@
+import { renderTagSystemForNotes } from './tag_system.js';
+
+export function getRaw(): string {
+  return `---
+model: gpt-5.2
+tools:
+  workspace-readFile:
+    type: builtin
+    name: workspace-readFile
+  workspace-edit:
+    type: builtin
+    name: workspace-edit
+  workspace-readdir:
+    type: builtin
+    name: workspace-readdir
+---
+# Task
+
+You are a note tagging agent. Given a batch of knowledge notes (People, Organizations, Projects, Topics), you will classify each note and prepend YAML frontmatter with categorized tags and Info attributes.
+
+# Instructions
+
+1. For each note file provided in the message, read its content carefully.
+2. Determine the note type from its folder path (People/, Organizations/, Projects/, Topics/).
+3. Classify the note using the Rowboat Tag System (Note Tags section) appended below.
+4. Extract attributes from the note's \`## Info\` section (or \`## About\` for Topics).
+5. Use \`workspace-edit\` to prepend YAML frontmatter to the file. The oldString should be the first line of the file (the \`# Title\` heading), and the newString should be the frontmatter followed by that same first line.
+6. If the note already has frontmatter (starts with \`---\`), skip it.
+
+# Frontmatter Format
+
+Tags are organized by **category** (not a flat list). Each tag category is a top-level YAML key. Use a plain string for single values, or a YAML list for multiple values.
+
+Info attributes from the \`## Info\` section are also included as top-level keys.
+
+\`\`\`yaml
+---
+relationship: customer
+relationship_sub: primary
+topic:
+  - sales
+  - fundraising
+source: email
+status: active
+action: action-required
+role: VP Engineering
+organization: Acme Corp
+email: sarah@acme.com
+first_met: "2024-06-15"
+last_seen: "2025-01-20"
+---
+\`\`\`
+
+## Tag category keys
+
+Use these exact keys for each tag category:
+
+| Category | Key | Single or multi | Example |
+|----------|-----|-----------------|---------|
+| Relationship | \`relationship\` | single | \`relationship: customer\` |
+| Relationship sub | \`relationship_sub\` | single or multi | \`relationship_sub: primary\` |
+| Topic | \`topic\` | single or multi | \`topic: sales\` or list |
+| Email type | \`email_type\` | single or multi | \`email_type: followup\` |
+| Action | \`action\` | single or multi | \`action: action-required\` |
+| Status | \`status\` | single | \`status: active\` |
+| Source | \`source\` | single or multi | \`source: email\` or list |
+
+**Rules:**
+- Use a plain string when there's only one value: \`topic: sales\`
+- Use a YAML list when there are multiple values:
+  \`\`\`yaml
+  topic:
+    - sales
+    - fundraising
+  \`\`\`
+- **Omit a category entirely** if no tags apply for it. Do not include empty keys.
+- Only use tag values from the Rowboat Tag System — do not invent new tags.
+
+# Info Attribute Extraction Rules
+
+Extract all \`**Key:** value\` fields from the \`## Info\` (or \`## About\`) section into YAML frontmatter keys:
+
+1. **Convert keys to snake_case**: e.g. \`**First met:**\` → \`first_met\`, \`**Last activity:**\` → \`last_activity\`, \`**Last seen:**\` → \`last_seen\`.
+2. **Strip wiki-link syntax**: \`[[Organizations/Acme Corp]]\` → \`Acme Corp\`. Extract just the display name (last path segment).
+3. **Skip blank/placeholder values**: If a field says "leave blank", is empty, or contains only template placeholders like \`{role}\`, omit it from the frontmatter.
+4. **Quote dates**: Wrap date values in quotes, e.g. \`first_met: "2024-06-15"\`.
+5. **Aliases as list**: If the value is comma-separated (like Aliases), store as a YAML list:
+   \`\`\`yaml
+   aliases:
+     - Sarah
+     - sarah@acme.com
+   \`\`\`
+
+**Per note type, extract these fields:**
+
+- **People**: role, organization, email, aliases, first_met, last_seen
+- **Organizations**: type, industry, relationship, domain, aliases, first_met, last_seen
+- **Projects**: type, status, started, last_activity
+- **Topics** (from \`## About\`): keywords, aliases, first_mentioned, last_mentioned
+
+Note: For Organizations, the Info \`**Relationship:**\` field is separate from the \`relationship\` tag category. Include both — the Info field as \`info_relationship\` and the tag as \`relationship\`.
+
+# Tag Selection Rules
+
+1. **Always include at least one relationship or topic tag** — every note must be classifiable.
+2. **Always include a source tag** — \`email\` or \`meeting\` based on what the note's Activity section shows.
+3. **Default status is \`active\`** for all new tags.
+4. **For People notes**, include:
+   - One primary relationship tag (e.g. \`customer\`, \`investor\`, \`prospect\`)
+   - Relationship sub-tags if applicable (e.g. \`primary\`, \`champion\`, \`former\`)
+   - Topic tags based on what you're working on together
+   - Source tags based on the Activity section
+   - Action tags if there are open items
+5. **For Organization notes**, include:
+   - One primary relationship tag
+   - Topic tags based on the relationship context
+   - Source tags
+6. **For Project notes**, include:
+   - Topic tags based on project type
+   - Source tags
+   - Action tags if there are open items
+7. **For Topic notes**, include:
+   - The relevant topic tag
+   - Source tags
+8. **Only use tags from the Rowboat Tag System** — do not invent new tags.
+9. Process all files in the batch. Do not skip any unless they already have frontmatter.
+
+---
+
+${renderTagSystemForNotes()}
+`;
+}
diff --git a/apps/x/packages/core/src/knowledge/note_tagging_state.ts b/apps/x/packages/core/src/knowledge/note_tagging_state.ts
new file mode 100644
index 00000000..ecfff8ea
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/note_tagging_state.ts
@@ -0,0 +1,48 @@
+import fs from 'fs';
+import path from 'path';
+import { WorkDir } from '../config/config.js';
+
+const STATE_FILE = path.join(WorkDir, 'note_tagging_state.json');
+
+export interface NoteTaggingState {
+    processedFiles: Record<string, { taggedAt: string }>;
+    lastRunTime: string;
+}
+
+export function loadNoteTaggingState(): NoteTaggingState {
+    if (fs.existsSync(STATE_FILE)) {
+        try {
+            return JSON.parse(fs.readFileSync(STATE_FILE, 'utf-8'));
+        } catch (error) {
+            console.error('Error loading note tagging state:', error);
+        }
+    }
+
+    return {
+        processedFiles: {},
+        lastRunTime: new Date(0).toISOString(),
+    };
+}
+
+export function saveNoteTaggingState(state: NoteTaggingState): void {
+    try {
+        fs.writeFileSync(STATE_FILE, JSON.stringify(state, null, 2));
+    } catch (error) {
+        console.error('Error saving note tagging state:', error);
+        throw error;
+    }
+}
+
+export function markNoteAsTagged(filePath: string, state: NoteTaggingState): void {
+    state.processedFiles[filePath] = {
+        taggedAt: new Date().toISOString(),
+    };
+}
+
+export function resetNoteTaggingState(): void {
+    const emptyState: NoteTaggingState = {
+        processedFiles: {},
+        lastRunTime: new Date().toISOString(),
+    };
+    saveNoteTaggingState(emptyState);
+}
diff --git a/apps/x/packages/core/src/knowledge/tag_notes.ts b/apps/x/packages/core/src/knowledge/tag_notes.ts
new file mode 100644
index 00000000..95934b03
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/tag_notes.ts
@@ -0,0 +1,274 @@
+import fs from 'fs';
+import path from 'path';
+import { WorkDir } from '../config/config.js';
+import { createRun, createMessage } from '../runs/runs.js';
+import { bus } from '../runs/bus.js';
+import { serviceLogger } from '../services/service_logger.js';
+import { limitEventItems } from './limit_event_items.js';
+import {
+    loadNoteTaggingState,
+    saveNoteTaggingState,
+    markNoteAsTagged,
+    type NoteTaggingState,
+} from './note_tagging_state.js';
+import { getNoteTypeDefinitions } from './note_system.js';
+
+const SYNC_INTERVAL_MS = 30 * 1000; // 30 seconds
+const BATCH_SIZE = 15;
+const NOTE_TAGGING_AGENT = 'note_tagging_agent';
+const KNOWLEDGE_DIR = path.join(WorkDir, 'knowledge');
+const MAX_CONTENT_LENGTH = 8000;
+
+/**
+ * Find knowledge notes that haven't been tagged yet
+ */
+function getUntaggedNotes(state: NoteTaggingState): string[] {
+    if (!fs.existsSync(KNOWLEDGE_DIR)) {
+        return [];
+    }
+
+    const untagged: string[] = [];
+    const noteFolders = getNoteTypeDefinitions().map(d => d.folder);
+
+    for (const folder of noteFolders) {
+        const folderPath = path.join(KNOWLEDGE_DIR, folder);
+        if (!fs.existsSync(folderPath)) {
+            continue;
+        }
+
+        const entries = fs.readdirSync(folderPath);
+        for (const entry of entries) {
+            const fullPath = path.join(folderPath, entry);
+            const stat = fs.statSync(fullPath);
+
+            if (!stat.isFile() || !entry.endsWith('.md')) {
+                continue;
+            }
+
+            // Skip if already tracked in state
+            if (state.processedFiles[fullPath]) {
+                continue;
+            }
+
+            // Skip if file already has frontmatter
+            try {
+                const content = fs.readFileSync(fullPath, 'utf-8');
+                if (content.startsWith('---')) {
+                    continue;
+                }
+            } catch {
+                continue;
+            }
+
+            untagged.push(fullPath);
+        }
+    }
+
+    return untagged;
+}
+
+/**
+ * Wait for a run to complete by listening for run-processing-end event
+ */
+async function waitForRunCompletion(runId: string): Promise<void> {
+    return new Promise(async (resolve) => {
+        const unsubscribe = await bus.subscribe('*', async (event) => {
+            if (event.type === 'run-processing-end' && event.runId === runId) {
+                unsubscribe();
+                resolve();
+            }
+        });
+    });
+}
+
+/**
+ * Tag a batch of note files using the tagging agent
+ */
+async function tagNoteBatch(
+    files: { path: string; content: string }[]
+): Promise<{ runId: string; filesEdited: Set<string> }> {
+    const run = await createRun({
+        agentId: NOTE_TAGGING_AGENT,
+    });
+
+    let message = `Tag the following ${files.length} knowledge notes by prepending YAML frontmatter with appropriate tags.\n\n`;
+    message += `**Important:** Use workspace-relative paths with workspace-edit (e.g. "knowledge/People/Sarah Chen.md", NOT absolute paths).\n\n`;
+
+    for (let i = 0; i < files.length; i++) {
+        const file = files[i];
+        const relativePath = path.relative(WorkDir, file.path);
+        const truncated = file.content.length > MAX_CONTENT_LENGTH
+            ? file.content.slice(0, MAX_CONTENT_LENGTH) + '\n\n[... content truncated, use workspace-readFile for full content ...]'
+            : file.content;
+
+        message += `## File ${i + 1}: ${relativePath}\n\n`;
+        message += truncated;
+        message += `\n\n---\n\n`;
+    }
+
+    const filesEdited = new Set<string>();
+
+    const unsubscribe = await bus.subscribe(run.id, async (event) => {
+        if (event.type !== 'tool-invocation') {
+            return;
+        }
+        if (event.toolName !== 'workspace-edit') {
+            return;
+        }
+        try {
+            const parsed = JSON.parse(event.input) as { path?: string };
+            if (typeof parsed.path === 'string') {
+                filesEdited.add(parsed.path);
+            }
+        } catch {
+            // ignore parse errors
+        }
+    });
+
+    await createMessage(run.id, message);
+    await waitForRunCompletion(run.id);
+    unsubscribe();
+
+    return { runId: run.id, filesEdited };
+}
+
+/**
+ * Process all untagged notes in batches
+ */
+async function processUntaggedNotes(): Promise<void> {
+    console.log('[NoteTagging] Checking for untagged notes...');
+
+    const state = loadNoteTaggingState();
+    const untagged = getUntaggedNotes(state);
+
+    if (untagged.length === 0) {
+        console.log('[NoteTagging] No untagged notes found');
+        return;
+    }
+
+    console.log(`[NoteTagging] Found ${untagged.length} untagged notes`);
+
+    const run = await serviceLogger.startRun({
+        service: 'note_tagging',
+        message: `Tagging ${untagged.length} note${untagged.length === 1 ? '' : 's'}`,
+        trigger: 'timer',
+    });
+
+    const relativeFiles = untagged.map(f => path.relative(WorkDir, f));
+    const limitedFiles = limitEventItems(relativeFiles);
+    await serviceLogger.log({
+        type: 'changes_identified',
+        service: run.service,
+        runId: run.runId,
+        level: 'info',
+        message: `Found ${untagged.length} untagged note${untagged.length === 1 ? '' : 's'}`,
+        counts: { notes: untagged.length },
+        items: limitedFiles.items,
+        truncated: limitedFiles.truncated,
+    });
+
+    const totalBatches = Math.ceil(untagged.length / BATCH_SIZE);
+    let totalEdited = 0;
+    let hadError = false;
+
+    for (let i = 0; i < untagged.length; i += BATCH_SIZE) {
+        const batchPaths = untagged.slice(i, i + BATCH_SIZE);
+        const batchNumber = Math.floor(i / BATCH_SIZE) + 1;
+
+        try {
+            const files: { path: string; content: string }[] = [];
+            for (const filePath of batchPaths) {
+                try {
+                    const content = fs.readFileSync(filePath, 'utf-8');
+                    files.push({ path: filePath, content });
+                } catch (error) {
+                    console.error(`[NoteTagging] Error reading ${filePath}:`, error);
+                }
+            }
+
+            if (files.length === 0) {
+                continue;
+            }
+
+            console.log(`[NoteTagging] Processing batch ${batchNumber}/${totalBatches} (${files.length} files)`);
+            await serviceLogger.log({
+                type: 'progress',
+                service: run.service,
+                runId: run.runId,
+                level: 'info',
+                message: `Processing batch ${batchNumber}/${totalBatches} (${files.length} files)`,
+                step: 'batch',
+                current: batchNumber,
+                total: totalBatches,
+                details: { filesInBatch: files.length },
+            });
+
+            const result = await tagNoteBatch(files);
+            totalEdited += result.filesEdited.size;
+
+            // Only mark files that were actually edited by the agent
+            for (const file of files) {
+                const relativePath = path.relative(WorkDir, file.path);
+                if (result.filesEdited.has(relativePath)) {
+                    markNoteAsTagged(file.path, state);
+                }
+            }
+
+            saveNoteTaggingState(state);
+            console.log(`[NoteTagging] Batch ${batchNumber}/${totalBatches} complete, ${result.filesEdited.size} files tagged`);
+        } catch (error) {
+            hadError = true;
+            console.error(`[NoteTagging] Error processing batch ${batchNumber}:`, error);
+            await serviceLogger.log({
+                type: 'error',
+                service: run.service,
+                runId: run.runId,
+                level: 'error',
+                message: `Error processing batch ${batchNumber}`,
+                error: error instanceof Error ? error.message : String(error),
+                context: { batchNumber },
+            });
+        }
+    }
+
+    state.lastRunTime = new Date().toISOString();
+    saveNoteTaggingState(state);
+
+    await serviceLogger.log({
+        type: 'run_complete',
+        service: run.service,
+        runId: run.runId,
+        level: hadError ? 'error' : 'info',
+        message: `Note tagging complete: ${totalEdited} notes tagged`,
+        durationMs: Date.now() - run.startedAt,
+        outcome: hadError ? 'error' : 'ok',
+        summary: {
+            totalNotes: untagged.length,
+            notesTagged: totalEdited,
+        },
+    });
+
+    console.log(`[NoteTagging] Done. ${totalEdited} notes tagged.`);
+}
+
+/**
+ * Main entry point - runs as independent polling service
+ */
+export async function init() {
+    console.log('[NoteTagging] Starting Note Tagging Service...');
+    console.log(`[NoteTagging] Will check for untagged notes every ${SYNC_INTERVAL_MS / 1000} seconds`);
+
+    // Initial run
+    await processUntaggedNotes();
+
+    // Periodic polling
+    while (true) {
+        await new Promise(resolve => setTimeout(resolve, SYNC_INTERVAL_MS));
+
+        try {
+            await processUntaggedNotes();
+        } catch (error) {
+            console.error('[NoteTagging] Error in main loop:', error);
+        }
+    }
+}
diff --git a/apps/x/packages/core/src/knowledge/tag_system.ts b/apps/x/packages/core/src/knowledge/tag_system.ts
new file mode 100644
index 00000000..01ac4c0e
--- /dev/null
+++ b/apps/x/packages/core/src/knowledge/tag_system.ts
@@ -0,0 +1,197 @@
+import path from "path";
+import fs from "fs";
+import { WorkDir } from "../config/config.js";
+
+export type TagApplicability = 'email' | 'notes' | 'both';
+
+export type TagType =
+  | 'relationship'
+  | 'relationship-sub'
+  | 'topic'
+  | 'email-type'
+  | 'filter'
+  | 'action'
+  | 'status'
+  | 'source';
+
+export interface TagDefinition {
+  tag: string;
+  type: TagType;
+  applicability: TagApplicability;
+  description: string;
+  example?: string;
+}
+
+// ── Default definitions (used to seed ~/.rowboat/config/tags.json) ──────────
+
+const DEFAULT_TAG_DEFINITIONS: TagDefinition[] = [
+  // ── Relationship (both) ──────────────────────────────────────────────
+  { tag: 'investor', type: 'relationship', applicability: 'both', description: 'Investors, VCs, or angels', example: 'Following up on our meeting — we\'d like to move forward with the Series A term sheet.' },
+  { tag: 'customer', type: 'relationship', applicability: 'both', description: 'Paying customers', example: 'We\'re seeing great results with Rowboat. Can we discuss expanding to more teams?' },
+  { tag: 'prospect', type: 'relationship', applicability: 'both', description: 'Potential customers', example: 'Thanks for the demo yesterday. We\'re interested in starting a pilot.' },
+  { tag: 'partner', type: 'relationship', applicability: 'both', description: 'Business partners', example: 'Let\'s discuss how we can promote the integration to both our user bases.' },
+  { tag: 'vendor', type: 'relationship', applicability: 'both', description: 'Service providers you work with', example: 'Here are the updated employment agreements you requested.' },
+  { tag: 'product', type: 'relationship', applicability: 'both', description: 'Products or services you use (automated)', example: 'Your AWS bill for January 2025 is now available.' },
+  { tag: 'candidate', type: 'relationship', applicability: 'both', description: 'Job applicants', example: 'Thanks for reaching out. I\'d love to learn more about the engineering role.' },
+  { tag: 'team', type: 'relationship', applicability: 'both', description: 'Internal team members', example: 'Here\'s the updated roadmap for Q2. Let\'s discuss in our sync.' },
+  { tag: 'advisor', type: 'relationship', applicability: 'both', description: 'Advisors, mentors, or board members', example: 'I\'ve reviewed the deck. Here are my thoughts on the GTM strategy.' },
+  { tag: 'personal', type: 'relationship', applicability: 'both', description: 'Family or friends', example: 'Are you coming to Thanksgiving this year? Let me know your travel dates.' },
+  { tag: 'press', type: 'relationship', applicability: 'both', description: 'Journalists or media', example: 'I\'m writing a piece on AI agents. Would you be available for an interview?' },
+  { tag: 'community', type: 'relationship', applicability: 'both', description: 'Users, peers, or open source contributors', example: 'Love what you\'re building with Rowboat. Here\'s a bug I found...' },
+  { tag: 'government', type: 'relationship', applicability: 'both', description: 'Government agencies', example: 'Your Delaware franchise tax is due by March 1, 2025.' },
+
+  // ── Relationship Sub-Tags (notes only) ───────────────────────────────
+  { tag: 'primary', type: 'relationship-sub', applicability: 'notes', description: 'Main contact or decision maker', example: 'Sarah Chen — VP Engineering, your main point of contact at Acme.' },
+  { tag: 'secondary', type: 'relationship-sub', applicability: 'notes', description: 'Supporting contact, involved but not the lead', example: 'David Kim — Engineer CC\'d on customer emails.' },
+  { tag: 'executive-assistant', type: 'relationship-sub', applicability: 'notes', description: 'EA or admin handling scheduling and logistics', example: 'Lisa — Sarah\'s EA who schedules all her meetings.' },
+  { tag: 'cc', type: 'relationship-sub', applicability: 'notes', description: 'Person who\'s CC\'d but not actively engaged', example: 'Manager looped in for visibility on deal.' },
+  { tag: 'referred-by', type: 'relationship-sub', applicability: 'notes', description: 'Person who made an introduction or referral', example: 'David Park — Investor who intro\'d you to Sarah.' },
+  { tag: 'former', type: 'relationship-sub', applicability: 'notes', description: 'Previously held this relationship, no longer active', example: 'John — Former customer who churned last year.' },
+  { tag: 'champion', type: 'relationship-sub', applicability: 'notes', description: 'Internal advocate pushing for you', example: 'Engineer who loves your product and is selling internally.' },
+  { tag: 'blocker', type: 'relationship-sub', applicability: 'notes', description: 'Person opposing or blocking progress', example: 'CFO resistant to spending on new tools.' },
+
+  // ── Topic (both) ─────────────────────────────────────────────────────
+  { tag: 'sales', type: 'topic', applicability: 'both', description: 'Sales conversations, deals, and revenue', example: 'Here\'s the pricing proposal we discussed. Let me know if you have questions.' },
+  { tag: 'support', type: 'topic', applicability: 'both', description: 'Help requests, issues, and customer support', example: 'We\'re seeing an error when trying to export. Can you help?' },
+  { tag: 'legal', type: 'topic', applicability: 'both', description: 'Contracts, terms, compliance, and legal matters', example: 'Legal has reviewed the MSA. Attached are our requested changes.' },
+  { tag: 'finance', type: 'topic', applicability: 'both', description: 'Money, invoices, payments, banking, and taxes', example: 'Your invoice #1234 for $5,000 is attached. Payment due in 30 days.' },
+  { tag: 'hiring', type: 'topic', applicability: 'both', description: 'Recruiting, interviews, and employment', example: 'We\'d like to move forward with a final round interview. Are you available Thursday?' },
+  { tag: 'fundraising', type: 'topic', applicability: 'both', description: 'Raising money and investor relations', example: 'Thanks for sending the deck. We\'d like to schedule a partner meeting.' },
+  { tag: 'travel', type: 'topic', applicability: 'both', description: 'Flights, hotels, trips, and travel logistics', example: 'Your flight to Tokyo on March 15 is confirmed. Confirmation #ABC123.' },
+  { tag: 'event', type: 'topic', applicability: 'both', description: 'Conferences, meetups, and gatherings', example: 'You\'re invited to speak at TechCrunch Disrupt. Can you confirm your availability?' },
+  { tag: 'shopping', type: 'topic', applicability: 'both', description: 'Purchases, orders, and returns', example: 'Your order #12345 has shipped. Track it here.' },
+  { tag: 'health', type: 'topic', applicability: 'both', description: 'Medical, wellness, and health-related matters', example: 'Your appointment with Dr. Smith is confirmed for Monday at 2pm.' },
+  { tag: 'learning', type: 'topic', applicability: 'both', description: 'Courses, education, and skill-building', example: 'Welcome to the Advanced Python course. Here\'s your access link.' },
+  { tag: 'research', type: 'topic', applicability: 'both', description: 'Research requests and information gathering', example: 'Here\'s the market analysis you requested on the AI agent space.' },
+
+  // ── Email Type ───────────────────────────────────────────────────────
+  { tag: 'intro', type: 'email-type', applicability: 'both', description: 'Warm introduction from someone you know', example: 'I\'d like to introduce you to Sarah Chen, VP Engineering at Acme.' },
+  { tag: 'followup', type: 'email-type', applicability: 'both', description: 'Following up on a previous conversation', example: 'Following up on our call last week. Have you had a chance to review the proposal?' },
+  { tag: 'scheduling', type: 'email-type', applicability: 'email', description: 'Meeting and calendar scheduling', example: 'Are you available for a call next Tuesday at 2pm?' },
+  { tag: 'cold-outreach', type: 'email-type', applicability: 'email', description: 'Unsolicited contact from someone you don\'t know', example: 'Hi, I noticed your company is growing fast. I\'d love to show you how we can help with...' },
+  { tag: 'newsletter', type: 'email-type', applicability: 'email', description: 'Newsletters, marketing emails, and subscriptions', example: 'This week in AI: The latest developments in agent frameworks...' },
+  { tag: 'notification', type: 'email-type', applicability: 'email', description: 'Automated alerts, receipts, and system notifications', example: 'Your password was changed successfully. If this wasn\'t you, contact support.' },
+
+  // ── Filter (email only) ──────────────────────────────────────────────
+  { tag: 'spam', type: 'filter', applicability: 'email', description: 'Junk and unwanted email', example: 'Congratulations! You\'ve won $1,000,000...' },
+  { tag: 'promotion', type: 'filter', applicability: 'email', description: 'Marketing offers and sales pitches', example: '50% off all items this weekend only!' },
+  { tag: 'social', type: 'filter', applicability: 'email', description: 'Social media notifications', example: 'John Smith commented on your post.' },
+  { tag: 'forums', type: 'filter', applicability: 'email', description: 'Mailing lists and group discussions', example: 'Re: [dev-list] Question about API design' },
+
+  // ── Action ───────────────────────────────────────────────────────────
+  { tag: 'action-required', type: 'action', applicability: 'both', description: 'Needs a response or action from you', example: 'Can you send me the pricing by Friday?' },
+  { tag: 'fyi', type: 'action', applicability: 'email', description: 'Informational only, no action needed', example: 'Just wanted to let you know the deal closed. Thanks for your help!' },
+  { tag: 'urgent', type: 'action', applicability: 'both', description: 'Time-sensitive, needs immediate attention', example: 'We need your signature on the contract by EOD today or we lose the deal.' },
+  { tag: 'waiting', type: 'action', applicability: 'both', description: 'Waiting on a response from them' },
+
+  // ── Status (email) ───────────────────────────────────────────────────
+  { tag: 'unread', type: 'status', applicability: 'email', description: 'Not yet processed' },
+  { tag: 'to-reply', type: 'status', applicability: 'email', description: 'Need to respond' },
+  { tag: 'done', type: 'status', applicability: 'email', description: 'Handled, can be archived' },
+
+  // ── Source (notes only) ──────────────────────────────────────────────
+  { tag: 'email', type: 'source', applicability: 'notes', description: 'Created or updated from email' },
+  { tag: 'meeting', type: 'source', applicability: 'notes', description: 'Created or updated from meeting transcript' },
+  { tag: 'browser', type: 'source', applicability: 'notes', description: 'Content captured from web browsing' },
+  { tag: 'web-search', type: 'source', applicability: 'notes', description: 'Information from web search' },
+  { tag: 'manual', type: 'source', applicability: 'notes', description: 'Manually entered by user' },
+  { tag: 'import', type: 'source', applicability: 'notes', description: 'Imported from another system' },
+
+  // ── Status (notes) ──────────────────────────────────────────────────
+  { tag: 'active', type: 'status', applicability: 'notes', description: 'Currently relevant, recent activity' },
+  { tag: 'archived', type: 'status', applicability: 'notes', description: 'No longer active, kept for reference' },
+  { tag: 'stale', type: 'status', applicability: 'notes', description: 'No activity in 60+ days, needs attention or archive' },
+];
+
+// ── Disk-backed config with mtime caching ──────────────────────────────────
+
+export const TAGS_CONFIG_PATH = path.join(WorkDir, "config", "tags.json");
+
+let cachedTagDefinitions: TagDefinition[] | null = null;
+let cachedMtimeMs: number | null = null;
+
+function ensureTagsConfigSync(): void {
+  if (!fs.existsSync(TAGS_CONFIG_PATH)) {
+    fs.writeFileSync(
+      TAGS_CONFIG_PATH,
+      JSON.stringify(DEFAULT_TAG_DEFINITIONS, null, 2) + "\n",
+      "utf8",
+    );
+  }
+}
+
+export function getTagDefinitions(): TagDefinition[] {
+  ensureTagsConfigSync();
+  try {
+    const stats = fs.statSync(TAGS_CONFIG_PATH);
+    if (cachedTagDefinitions && cachedMtimeMs === stats.mtimeMs) {
+      return cachedTagDefinitions;
+    }
+    const content = fs.readFileSync(TAGS_CONFIG_PATH, "utf8");
+    cachedTagDefinitions = JSON.parse(content);
+    cachedMtimeMs = stats.mtimeMs;
+    return cachedTagDefinitions!;
+  } catch {
+    cachedTagDefinitions = null;
+    cachedMtimeMs = null;
+    return DEFAULT_TAG_DEFINITIONS;
+  }
+}
+
+// ── Render helpers ───────────────────────────────────────────────────────
+
+const TYPE_ORDER: TagType[] = [
+  'relationship', 'relationship-sub', 'topic', 'email-type',
+  'filter', 'action', 'status', 'source',
+];
+
+const TYPE_LABELS: Record<TagType, string> = {
+  'relationship': 'Relationship',
+  'relationship-sub': 'Relationship Sub-Tags',
+  'topic': 'Topic',
+  'email-type': 'Email Type',
+  'filter': 'Filter',
+  'action': 'Action',
+  'status': 'Status',
+  'source': 'Source',
+};
+
+function renderTagGroups(tags: TagDefinition[]): string {
+  const groups = new Map<TagType, TagDefinition[]>();
+  for (const tag of tags) {
+    const list = groups.get(tag.type) ?? [];
+    list.push(tag);
+    groups.set(tag.type, list);
+  }
+
+  const sections: string[] = [];
+  for (const type of TYPE_ORDER) {
+    const group = groups.get(type);
+    if (!group || group.length === 0) continue;
+
+    const label = TYPE_LABELS[type];
+    const rows = group.map(t => {
+      const example = t.example ?? '';
+      return `| ${t.tag} | ${t.description} | ${example} |`;
+    });
+
+    sections.push(
+      `## ${label}\n\n` +
+      `| Tag | Description | Example |\n` +
+      `|-----|-------------|---------|\n` +
+      rows.join('\n'),
+    );
+  }
+
+  return `# Tag System Reference\n\n${sections.join('\n\n')}`;
+}
+
+export function renderTagSystemForNotes(): string {
+  const tags = getTagDefinitions().filter(t => t.applicability !== 'email');
+  return renderTagGroups(tags);
+}
+
+export function renderTagSystemForEmails(): string {
+  const tags = getTagDefinitions().filter(t => t.applicability !== 'notes');
+  return renderTagGroups(tags);
+}
diff --git a/apps/x/packages/shared/src/service-events.ts b/apps/x/packages/shared/src/service-events.ts
index d214472c..807bc063 100644
--- a/apps/x/packages/shared/src/service-events.ts
+++ b/apps/x/packages/shared/src/service-events.ts
@@ -7,6 +7,8 @@ export const ServiceName = z.enum([
   'fireflies',
   'granola',
   'voice_memo',
+  'email_labeling',
+  'note_tagging',
 ]);
 
 const ServiceEventBase = z.object({