Merge pull request #533 from rowboatlabs/coding2

Coding2

apps/x/packages/core/src/agents/runtime.ts

@@ -176,6 +176,7 @@ export class AgentRuntime implements IAgentRuntime {
                         modelConfigRepo: this.modelConfigRepo,
                         signal,
                         abortRegistry: this.abortRegistry,
+                        bus: this.bus,
                     })) {
                         eventCount++;
                         if (event.type !== "llm-stream-event") {
@@ -868,6 +869,7 @@ export async function* streamAgent({
     modelConfigRepo,
     signal,
     abortRegistry,
+    bus,
 }: {
     state: AgentState,
     idGenerator: IMonotonicallyIncreasingIdGenerator;
@@ -876,6 +878,7 @@ export async function* streamAgent({
     modelConfigRepo: IModelConfigRepo;
     signal: AbortSignal;
     abortRegistry: IAbortRegistry;
+    bus: IBus;
 }): AsyncGenerator<z.infer<typeof RunEvent>, void, unknown> {
     const logger = new PrefixLogger(`run-${runId}-${state.agentName}`);
 
@@ -985,11 +988,12 @@ export async function* streamAgent({
                         state: subflowState,
                         idGenerator,
                         runId,
-                        messageQueue,
-                        modelConfigRepo,
-                        signal,
-                        abortRegistry,
-                    })) {
+                    messageQueue,
+                    modelConfigRepo,
+                    signal,
+                    abortRegistry,
+                    bus,
+                })) {
                         yield* processEvent({
                             ...event,
                             subflow: [toolCallId, ...event.subflow],
@@ -998,9 +1002,15 @@ export async function* streamAgent({
                     if (!subflowState.getPendingAskHumans().length && !subflowState.getPendingPermissions().length) {
                         result = subflowState.finalResponse();
                     }
-                } else {
-                    result = await execTool(agent.tools![toolCall.toolName], toolCall.arguments, { runId, signal, abortRegistry });
-                }
+            } else {
+                result = await execTool(agent.tools![toolCall.toolName], toolCall.arguments, {
+                    runId,
+                    toolCallId,
+                    signal,
+                    abortRegistry,
+                    publish: (event) => bus.publish(event),
+                });
+            }
             } catch (error) {
                 if ((error instanceof Error && error.name === "AbortError") || signal.aborted) {
                     throw error;

apps/x/packages/core/src/application/assistant/instructions.ts

@@ -80,6 +80,8 @@ ${thirdPartyBlock}**Meeting Prep:** When users ask you to prepare for a meeting,
 
 **Document Collaboration:** When users ask you to work on a document, collaborate on writing, create a new document, edit/refine existing notes, or say things like "let's work on [X]", "help me write [X]", "create a doc for [X]", or "let's draft [X]", you MUST load the \`doc-collab\` skill first. This is required for any document creation or editing task. The skill provides structured guidance for creating, editing, and refining documents in the knowledge base.
 
+**Code with Agents:** When users ask you to write code, build a project, create a script, fix a bug, or do any software development task, load the \`code-with-agents\` skill first. It provides guidance for delegating coding work to Claude Code or Codex via acpx.
+
 **App Control:** When users ask you to open notes, show the bases or graph view, filter or search notes, or manage saved views, load the \`app-navigation\` skill first. It provides structured guidance for navigating the app UI and controlling the knowledge base view.
 
 **Tracks (Auto-Updating Note Blocks):** When users ask you to **track**, **monitor**, **watch**, or **keep an eye on** something in a note — or say things like "every morning tell me X", "show the current Y in this note", "pin live updates of Z here" — load the \`tracks\` skill first. Also load it when a user presses Cmd+K with a note open and requests auto-refreshing content at the cursor. Track blocks are YAML-fenced scheduled blocks whose output is rewritten on each run — useful for weather, news, prices, status pages, and personal dashboards.

apps/x/packages/core/src/application/assistant/skills/code-with-agents/skill.ts

@@ -0,0 +1,90 @@
+export const skill = String.raw`
+# Code with Agents Skill
+
+Use this skill when the user asks you to write code, build a project, create scripts, fix bugs, or do any software development task that should be delegated to a coding agent (Claude Code or Codex).
+
+## Important: delegate ALL coding work
+
+Once the user has chosen to use Claude Code or Codex, you MUST delegate ALL code-related tasks to the coding agent. This includes:
+- Writing, editing, or refactoring code
+- Reading, summarizing, or explaining code
+- Debugging and fixing bugs
+- Running tests or build commands
+- Exploring project structure
+- Any other task that involves interacting with a codebase
+
+Do NOT attempt to do any of these yourself — no reading files, no running commands, no writing code. You are the coordinator; the coding agent does the work. Your job is to translate the user's request into a clear prompt and pass it to the agent.
+
+## Prerequisites
+
+The user must have one of the following installed on their machine:
+- **Claude Code** — https://claude.ai/code
+- **Codex** — https://codex.openai.com
+
+These are external tools that you cannot install for the user.
+
+## Workflow
+
+### Step 1: Gather requirements
+
+Before running anything, confirm the following with the user:
+
+1. **Working directory** — Ask which folder the code should be written in, unless the user has already specified it. Example: "Which folder should I work in?"
+2. **Agent choice** — Ask whether to use **Claude Code** or **Codex**. Mention that the chosen agent must already be installed on their machine.
+
+### Step 2: Confirm execution plan
+
+Once you know the folder and agent, tell the user:
+
+> I'll use [Claude Code / Codex] to [description of the task] in \`[folder]\`. Permission requests from the coding agent itself (file writes, command execution, etc.) will be automatically approved once it starts. Wait for the user's confirmation before you execute anything.
+
+### Step 3: Execute with acpx
+
+Use the \`executeCommand\` tool to run the coding agent via acpx. The command format is:
+
+**For Claude Code:**
+` + "`" + `
+npx acpx@latest --approve-all --cwd <folder> claude exec "<prompt>"
+` + "`" + `
+
+**For Codex:**
+` + "`" + `
+npx acpx@latest --approve-all --cwd <folder> codex exec "<prompt>"
+` + "`" + `
+
+### Critical: flag order
+
+The \`--approve-all\` and \`--cwd\` flags are global flags and MUST come before the agent name (\`claude\` or \`codex\`). This is the correct order:
+
+` + "`" + `
+npx acpx@latest [global flags] <agent> exec "<prompt>"
+` + "`" + `
+
+**Correct:**
+` + "`" + `
+npx acpx@latest --approve-all --cwd ~/projects/myapp claude exec "fix the bug"
+` + "`" + `
+
+**Wrong (will fail):**
+` + "`" + `
+npx acpx@latest claude --approve-all exec "fix the bug"
+` + "`" + `
+
+### Writing good prompts
+
+When constructing the prompt for the coding agent:
+- Be specific and detailed about what to build or fix
+- Include file names, function signatures, and expected behavior
+- Mention any constraints (language, framework, style)
+- If the user gave you a short request, expand it into a clear, actionable prompt for the agent
+
+### Step 4: Report results
+
+After the command finishes, look for the summary that the coding agent produced at the end of its output and pass that along to the user as-is. Do not rewrite or add to it. Only add your own explanation if the command failed or the exit code is non-zero.
+
+Do NOT use file reference blocks (e.g. \`\`\`file:path/to/file\`\`\`) when mentioning code files — they may not open correctly. Just refer to file paths as plain text.
+
+- If the exit code is 5, it means permissions were denied — this should not happen with \`--approve-all\`, but if it does, let the user know
+`;
+
+export default skill;

apps/x/packages/core/src/application/assistant/skills/index.ts

@@ -11,6 +11,7 @@ import createPresentationsSkill from "./create-presentations/skill.js";
 
 import appNavigationSkill from "./app-navigation/skill.js";
 import browserControlSkill from "./browser-control/skill.js";
+import codeWithAgentsSkill from "./code-with-agents/skill.js";
 import composioIntegrationSkill from "./composio-integration/skill.js";
 import tracksSkill from "./tracks/skill.js";
 import notifyUserSkill from "./notify-user/skill.js";
@@ -94,6 +95,12 @@ const definitions: SkillDefinition[] = [
     summary: "Navigate the app UI - open notes, switch views, filter/search the knowledge base, and manage saved views.",
     content: appNavigationSkill,
   },
+  {
+    id: "code-with-agents",
+    title: "Code with Agents",
+    summary: "Write code, build projects, create scripts, or fix bugs by delegating to Claude Code or Codex via acpx.",
+    content: codeWithAgentsSkill,
+  },
   {
     id: "tracks",
     title: "Tracks",

apps/x/packages/core/src/application/lib/builtin-tools.ts

@@ -969,6 +969,16 @@ export const BuiltinTools: z.infer<typeof BuiltinToolsSchema> = {
                     const { promise, process: proc } = executeCommandAbortable(command, {
                         cwd: workingDir,
                         signal: ctx.signal,
+                        onData: (chunk: string) => {
+                            ctx.publish({
+                                runId: ctx.runId,
+                                type: "tool-output-stream",
+                                toolCallId: ctx.toolCallId,
+                                toolName: "executeCommand",
+                                output: chunk,
+                                subflow: [],
+                            });
+                        },
                     });
 
                     // Register process with abort registry for force-kill

apps/x/packages/core/src/application/lib/command-executor.ts

@@ -143,6 +143,7 @@ export function executeCommandAbortable(
     timeout?: number;
     maxBuffer?: number;
     signal?: AbortSignal;
+    onData?: (chunk: string) => void;
   }
 ): { promise: Promise<AbortableCommandResult>; process: ChildProcess } {
   // Check if already aborted before spawning
@@ -177,16 +178,20 @@ export function executeCommandAbortable(
 
     // Collect output
     proc.stdout?.on('data', (chunk: Buffer) => {
+      const text = chunk.toString();
       const maxBuffer = options?.maxBuffer || 1024 * 1024;
       if (stdout.length < maxBuffer) {
-        stdout += chunk.toString();
+        stdout += text;
       }
+      options?.onData?.(text);
     });
     proc.stderr?.on('data', (chunk: Buffer) => {
+      const text = chunk.toString();
       const maxBuffer = options?.maxBuffer || 1024 * 1024;
       if (stderr.length < maxBuffer) {
-        stderr += chunk.toString();
+        stderr += text;
       }
+      options?.onData?.(text);
     });
 
     // Abort handler

apps/x/packages/core/src/application/lib/exec-tool.ts

@@ -1,4 +1,5 @@
 import { ToolAttachment } from "@x/shared/dist/agent.js";
+import { RunEvent } from "@x/shared/dist/runs.js";
 import { z } from "zod";
 import { BuiltinTools } from "./builtin-tools.js";
 import { executeTool } from "../../mcp/mcp.js";
@@ -9,8 +10,10 @@ import { IAbortRegistry } from "../../runs/abort-registry.js";
  */
 export interface ToolContext {
     runId: string;
+    toolCallId: string;
     signal: AbortSignal;
     abortRegistry: IAbortRegistry;
+    publish: (event: z.infer<typeof RunEvent>) => Promise<void>;
 }
 
 async function execMcpTool(agentTool: z.infer<typeof ToolAttachment> & { type: "mcp" }, input: Record<string, unknown>): Promise<unknown> {
@@ -34,4 +37,4 @@ export async function execTool(agentTool: z.infer<typeof ToolAttachment>, input:
             return builtinTool.execute(input, ctx);
         }
     }
-}
+}