Polish ktx setup-agents output: hint, summaries, outro

This commit is contained in:
Andrey Avtomonov 2026-05-19 14:12:18 +02:00
parent ddabe517e3
commit 0cbf8121d9
4 changed files with 1151 additions and 87 deletions

View file

@ -0,0 +1,728 @@
# KTX Setup Agents Output Polish Implementation Plan
> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
**Goal:** Make `ktx setup --agents` print the polished interactive hint,
per-install inline summaries, the required next-actions note, and the final
`All set.` outro described by the May 19 output-polish spec.
**Architecture:** Keep the change centered in
`packages/cli/src/setup-agents.ts` and preserve the existing
`withMultiselectNavigation` helper for other setup flows. Replace the old
boxed install-summary string with structured summary entries, then render those
entries with Clack `log.step` for real TTY output and plain `io.stdout` text in
tests or scripted output. The current uncommitted Claude Desktop split-ZIP
changes are treated as the baseline and must not be reverted.
**Tech Stack:** TypeScript, Vitest, Node stream types, `@clack/prompts`,
pnpm workspace commands.
---
## Current audit
Original spec:
`docs/superpowers/specs/2026-05-19-ktx-setup-agents-output-polish-design.md`
Current uncommitted files:
- `README.md`
- `docs-site/content/docs/integrations/agent-clients.mdx`
- `packages/cli/src/setup-agents.test.ts`
- `packages/cli/src/setup-agents.ts`
Observed implementation status:
- The branch currently contains unrelated Claude Desktop split-ZIP work:
`ktx-analytics.zip` and optional `ktx.zip` replace the old combined
`ktx-skills.zip`. Keep that work intact.
- `packages/cli/src/setup-agents.ts` still imports and uses
`withMultiselectNavigation` for the agents multiselect prompt.
- `runKtxSetupAgentsStep` does not emit
`Space to select, Enter to confirm, Esc to go back.` before interactive
prompts.
- `formatInstallSummary` still returns one boxed string containing
`KTX project`, `Installed agents`, and nested indentation.
- `runKtxSetupAgentsStep` still prints the install summary with
`setupUi.note(..., 'Agent integration complete', io)`.
- `runKtxSetupAgentsStep` does not emit `outro('All set.')` after the
"Required before using agents" note.
- Tests still assert the old long multiselect message and old boxed install
summary.
V1-blocking gaps:
- Move the multiselect navigation hint out of the question text and print it
once before interactive prompts.
- Render install summaries as inline `log.step` blocks, one block per install.
- Keep the next-actions note boxed and emit `All set.` only when next-actions
are shown.
- Update setup-agents tests to cover the new output contract.
Non-blocking gaps:
- The docs updates for Claude Desktop split ZIPs are outside this output-polish
spec and are already present in the current working tree.
- `setup --sources` and `setup --databases` still use
`withMultiselectNavigation`; the spec explicitly defers those flows.
- There is no docs-site update for the output shape because no public docs page
currently promises the old boxed install-summary layout.
## File structure
Modify:
- `packages/cli/src/setup-agents.ts`
- Remove the agents-flow dependency on `withMultiselectNavigation`.
- Add tiny output helpers that use Clack `log.info`, `log.step`, and `outro`
when `io.stdout` is a writable TTY and fall back to plain writes otherwise.
- Rename `formatInstallSummary` to `formatInstallSummaryLines` and return
structured entries.
- Render each summary entry with `log.step`.
- Emit the final outro only when next-actions are printed.
- `packages/cli/src/setup-agents.test.ts`
- Import `formatInstallSummaryLines`.
- Update the four output-polish tests named in the spec.
- Add or update assertions for the single hint line and final outro.
Do not modify:
- `packages/cli/src/prompt-navigation.ts`
- `packages/cli/src/prompt-navigation.test.ts`
- `README.md`
- `docs-site/content/docs/integrations/agent-clients.mdx`
## Task 1: Move the interactive multiselect hint out of the prompt
**Files:**
- Modify: `packages/cli/src/setup-agents.test.ts`
- Modify: `packages/cli/src/setup-agents.ts`
- [ ] **Step 1: Write the failing prompt-output test**
In `packages/cli/src/setup-agents.test.ts`, replace the existing test named
`explains how to select multiple agent targets in interactive mode` with:
```typescript
it('prints one navigation hint before interactive agent target prompts', async () => {
const io = makeIo();
const prompts = {
select: vi.fn(async () => 'mcp-cli'),
multiselect: vi.fn(async () => ['back']),
cancel: vi.fn(),
};
await expect(
runKtxSetupAgentsStep(
{
projectDir: tempDir,
inputMode: 'auto',
yes: false,
agents: true,
scope: 'project',
mode: 'mcp-cli',
skipAgents: false,
},
io.io,
{ prompts },
),
).resolves.toEqual({ status: 'back', projectDir: tempDir });
expect(io.stdout()).toContain('Space to select, Enter to confirm, Esc to go back.');
expect(io.stdout().match(/Space to select/g)).toHaveLength(1);
expect(prompts.multiselect).toHaveBeenCalledWith(
expect.objectContaining({
message: 'Which agent targets should KTX install?',
}),
);
});
```
- [ ] **Step 2: Run the focused test and verify it fails**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "prints one navigation hint before interactive agent target prompts"
```
Expected: FAIL because the current implementation embeds the long navigation
hint in the multiselect message and does not write the short hint to
`io.stdout`.
- [ ] **Step 3: Add Clack output helpers for setup-agents**
In `packages/cli/src/setup-agents.ts`, replace the current top imports:
```typescript
import { existsSync } from 'node:fs';
import { chmod, mkdir, readFile, rm, writeFile } from 'node:fs/promises';
import { dirname, join, relative, resolve } from 'node:path';
import { fileURLToPath } from 'node:url';
```
with:
```typescript
import { existsSync } from 'node:fs';
import { chmod, mkdir, readFile, rm, writeFile } from 'node:fs/promises';
import { dirname, join, relative, resolve } from 'node:path';
import type { Writable } from 'node:stream';
import { fileURLToPath } from 'node:url';
import { log, outro } from '@clack/prompts';
```
Remove this import:
```typescript
import { withMultiselectNavigation } from './prompt-navigation.js';
```
Add these helpers after the `KtxCliLauncher` interface:
```typescript
function isWritableTtyOutput(output: KtxCliIo['stdout']): output is KtxCliIo['stdout'] & Writable {
return (
output.isTTY === true &&
typeof (output as { on?: unknown }).on === 'function' &&
typeof (output as { columns?: unknown }).columns !== 'undefined'
);
}
function writeSetupInfo(io: KtxCliIo, message: string): void {
if (isWritableTtyOutput(io.stdout)) {
log.info(message, { output: io.stdout });
return;
}
io.stdout.write(`${message}\n`);
}
function writeSetupStep(io: KtxCliIo, message: string): void {
if (isWritableTtyOutput(io.stdout)) {
log.step(message, { output: io.stdout });
return;
}
io.stdout.write(`\n${message}\n`);
}
function writeSetupOutro(io: KtxCliIo, message: string): void {
if (isWritableTtyOutput(io.stdout)) {
outro(message, { output: io.stdout });
return;
}
io.stdout.write(`\n${message}\n`);
}
```
- [ ] **Step 4: Print the short hint only before interactive prompts**
In `runKtxSetupAgentsStep`, immediately after:
```typescript
const prompts = deps.prompts ?? createPromptAdapter();
```
add:
```typescript
if (args.inputMode === 'auto' && args.target === undefined) {
writeSetupInfo(io, 'Space to select, Enter to confirm, Esc to go back.');
}
```
In the multiselect call, replace:
```typescript
message: withMultiselectNavigation('Which agent targets should KTX install?'),
```
with:
```typescript
message: 'Which agent targets should KTX install?',
```
- [ ] **Step 5: Run the focused test and verify it passes**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "prints one navigation hint before interactive agent target prompts"
```
Expected: PASS.
## Task 2: Return structured install-summary entries
**Files:**
- Modify: `packages/cli/src/setup-agents.test.ts`
- Modify: `packages/cli/src/setup-agents.ts`
- [ ] **Step 1: Update the summary formatter import in tests**
In `packages/cli/src/setup-agents.test.ts`, replace:
```typescript
formatInstallSummary,
```
with:
```typescript
formatInstallSummaryLines,
```
- [ ] **Step 2: Write failing structured formatter tests**
In `packages/cli/src/setup-agents.test.ts`, replace the test named
`formats summary with explicit project-scoped config paths` with:
```typescript
it('formats summary with explicit project-scoped config paths', () => {
const summary = formatInstallSummaryLines(
[{ target: 'cursor', scope: 'project', mode: 'mcp-cli' }],
[
{ kind: 'file', path: join(tempDir, '.cursor/rules/ktx-analytics.mdc'), role: 'analytics-skill' },
{ kind: 'file', path: join(tempDir, '.cursor/rules/ktx.mdc') },
{ kind: 'json-key', path: join(tempDir, '.cursor/mcp.json'), jsonPath: ['mcpServers', 'ktx'] },
],
tempDir,
);
expect(summary).toEqual([
{
title: 'Cursor · Project scope',
lines: [
join(tempDir, '.cursor/mcp.json'),
'Requires MCP to be started.',
'Cursor rules installed.',
],
},
]);
});
```
Replace the test named `formats summary with multiple agent targets` with:
```typescript
it('formats summary with multiple agent targets', () => {
const summary = formatInstallSummaryLines(
[
{ target: 'claude-code', scope: 'project', mode: 'mcp-cli' },
{ target: 'codex', scope: 'project', mode: 'mcp-cli' },
],
[
{ kind: 'file', path: join(tempDir, '.claude/skills/ktx-analytics/SKILL.md'), role: 'analytics-skill' },
{ kind: 'file', path: join(tempDir, '.claude/skills/ktx/SKILL.md'), role: 'skill' },
{ kind: 'file', path: join(tempDir, '.claude/rules/ktx.md'), role: 'rule' },
{ kind: 'json-key', path: join(tempDir, '.mcp.json'), jsonPath: ['mcpServers', 'ktx'] },
{ kind: 'file', path: join(tempDir, '.agents/skills/ktx-analytics/SKILL.md'), role: 'analytics-skill' },
{ kind: 'file', path: join(tempDir, '.agents/skills/ktx/SKILL.md'), role: 'skill' },
{ kind: 'file', path: join(tempDir, '.codex/instructions/ktx.md'), role: 'rule' },
],
tempDir,
);
expect(summary).toEqual([
{
title: 'Claude Code · Project scope',
lines: [
join(tempDir, '.mcp.json'),
'Requires MCP to be started.',
'Analytics skill installed.',
'Admin CLI skill installed.',
],
},
{
title: 'Codex · Project scope',
lines: [
'Add the snippet shown below to ~/.codex/config.toml.',
'Requires MCP to be started.',
'Codex guidance installed.',
],
},
]);
});
```
- [ ] **Step 3: Run the focused formatter tests and verify they fail**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "formats summary"
```
Expected: FAIL because `formatInstallSummaryLines` is not exported and the
current formatter returns one string.
- [ ] **Step 4: Replace the string summary formatter with structured entries**
In `packages/cli/src/setup-agents.ts`, add this exported interface immediately
before the current `formatInstallSummary` function:
```typescript
export interface InstallSummaryEntry {
title: string;
lines: string[];
}
```
Replace the full `export function formatInstallSummary(...)` implementation
with:
```typescript
function formatInlinePath(path: string): string {
const home = process.env.HOME;
if (!home) return path;
const resolvedHome = resolve(home);
if (path === resolvedHome) return '~';
if (path.startsWith(`${resolvedHome}/`)) {
return `~/${relative(resolvedHome, path)}`;
}
return path;
}
export function formatInstallSummaryLines(
installs: Array<{ target: KtxAgentTarget; scope: KtxAgentScope; mode: KtxAgentInstallMode }>,
entries: InstallEntry[],
projectDir: string,
): InstallSummaryEntry[] {
const entriesByTarget = new Map<KtxAgentTarget, InstallEntry[]>();
for (const install of installs) {
const plannedFilePaths = new Set(
plannedKtxAgentFiles({ projectDir, ...install })
.filter((entry) => entry.kind === 'file')
.map((entry) => entry.path),
);
entriesByTarget.set(
install.target,
entries.filter((entry) => entry.kind === 'file' && plannedFilePaths.has(entry.path)),
);
}
const mcpEntriesByTarget = new Map<KtxAgentTarget, InstallEntry[]>();
for (const install of installs) {
const plannedMcpKeys = new Set(plannedMcpJsonEntries({ projectDir, ...install }).map(entryKey));
mcpEntriesByTarget.set(
install.target,
entries.filter((entry) => entry.kind === 'json-key' && plannedMcpKeys.has(entryKey(entry))),
);
}
return installs.map((install) => {
const targetEntries = entriesByTarget.get(install.target) ?? [];
const mcpEntry = mcpEntriesByTarget
.get(install.target)
?.find((entry): entry is Extract<InstallEntry, { kind: 'json-key' }> => entry.kind === 'json-key');
const lines: string[] = [];
if (mcpEntry) {
lines.push(formatInlinePath(mcpEntry.path));
} else if (install.target !== 'claude-desktop') {
lines.push(manualMcpConfigInstruction(install.target, install.scope));
}
if (targetUsesHttpMcpDaemon(install.target)) {
lines.push('Requires MCP to be started.');
}
const hasAnalytics = hasEntryRole(targetEntries, 'analytics-skill');
const hasAdmin = hasAdminCliEntries(targetEntries);
const claudeDesktopSkillBundles = targetEntries.filter(
(entry): entry is Extract<InstallEntry, { kind: 'file' }> =>
entry.kind === 'file' && entry.role === 'claude-desktop-skill-bundle',
);
if (install.target === 'claude-code') {
if (hasAnalytics) {
lines.push('Analytics skill installed.');
}
if (hasAdmin) {
lines.push('Admin CLI skill installed.');
}
} else if (install.target === 'claude-desktop') {
if (claudeDesktopSkillBundles.length > 0) {
lines.push('Skill bundles:');
for (const bundle of claudeDesktopSkillBundles) {
lines.push(` ${bundle.path}`);
}
}
} else if (hasAnalytics || hasAdmin) {
lines.push(`${guidanceInstallLine(install.target)}.`);
}
if (hasEntryRole(targetEntries, 'launcher')) {
lines.push('Starts KTX over stdio from Claude Desktop.');
}
return {
title: `${targetDisplayName(install.target)} · ${scopeDisplayName(install.scope)}`,
lines,
};
});
}
```
- [ ] **Step 5: Run the focused formatter tests and verify they pass**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "formats summary"
```
Expected: PASS.
## Task 3: Render inline install summaries with `log.step`
**Files:**
- Modify: `packages/cli/src/setup-agents.test.ts`
- Modify: `packages/cli/src/setup-agents.ts`
- [ ] **Step 1: Write failing install-summary output assertions**
In `packages/cli/src/setup-agents.test.ts`, replace the body of the test named
`prints per-agent install summary after successful installation` with:
```typescript
const io = makeIo();
await runKtxSetupAgentsStep(
{
projectDir: tempDir,
inputMode: 'disabled',
yes: true,
agents: true,
target: 'claude-code',
scope: 'project',
mode: 'mcp-cli',
skipAgents: false,
},
io.io,
);
const output = io.stdout();
expect(output).toContain('Claude Code · Project scope');
expect(output).toContain(join(tempDir, '.mcp.json'));
expect(output).toContain('Requires MCP to be started.');
expect(output).toContain('Analytics skill installed.');
expect(output).toContain('Admin CLI skill installed.');
expect(output).not.toContain('Agent integration complete');
expect(output).not.toContain(`KTX project\n ${tempDir}`);
expect(output).not.toContain('Installed agents');
expect(output).not.toContain('.claude/skills/ktx-analytics/SKILL.md');
expect(output).not.toContain('.claude/skills/ktx/SKILL.md');
expect(output).not.toContain('.claude/rules/ktx.md');
```
In the test named `can return agent next actions without printing them`, replace:
```typescript
expect(io.stdout()).toContain('Agent integration complete');
expect(io.stdout()).not.toContain('Required before using agents');
```
with:
```typescript
expect(io.stdout()).toContain('Claude Code · Project scope');
expect(io.stdout()).not.toContain('Agent integration complete');
expect(io.stdout()).not.toContain('Required before using agents');
expect(io.stdout()).not.toContain('All set.');
```
- [ ] **Step 2: Run the focused output tests and verify they fail**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "install summary|without printing"
```
Expected: FAIL because `Agent integration complete` is still printed and the
new inline summary heading is absent.
- [ ] **Step 3: Render each structured summary as an inline step**
In `runKtxSetupAgentsStep`, replace:
```typescript
setupUi.note(
formatInstallSummary(installs, entries, args.projectDir),
'Agent integration complete',
io,
);
```
with:
```typescript
for (const summary of formatInstallSummaryLines(installs, entries, args.projectDir)) {
writeSetupStep(
io,
summary.lines.length > 0 ? `${summary.title}\n${summary.lines.join('\n')}` : summary.title,
);
}
```
- [ ] **Step 4: Run the focused output tests and verify they pass**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "install summary|without printing"
```
Expected: PASS.
## Task 4: Emit the final outro after next actions
**Files:**
- Modify: `packages/cli/src/setup-agents.test.ts`
- Modify: `packages/cli/src/setup-agents.ts`
- [ ] **Step 1: Add failing outro assertions**
In `packages/cli/src/setup-agents.test.ts`, in the test named
`prints standalone agent next actions after successful installation`, add this
assertion after the existing next-actions stdout assertions:
```typescript
expect(io.stdout()).toContain('All set.');
```
In the test named `prints one target-aware next actions block for mixed agent
targets`, add this assertion near the other stdout assertions:
```typescript
expect(output).toContain('All set.');
```
- [ ] **Step 2: Run the focused next-actions tests and verify they fail**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "next actions|All set|without printing"
```
Expected: FAIL because no outro is emitted.
- [ ] **Step 3: Emit `All set.` only when next-actions are shown**
In `runKtxSetupAgentsStep`, replace:
```typescript
if (args.showNextActions !== false) {
setupUi.note(nextActions, 'Required before using agents', io, { format: (line) => line });
}
```
with:
```typescript
if (args.showNextActions !== false) {
setupUi.note(nextActions, 'Required before using agents', io, { format: (line) => line });
writeSetupOutro(io, 'All set.');
}
```
- [ ] **Step 4: Run the focused next-actions tests and verify they pass**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts -t "next actions|All set|without printing"
```
Expected: PASS.
## Task 5: Verify the full CLI package surface
**Files:**
- Verify: `packages/cli/src/setup-agents.ts`
- Verify: `packages/cli/src/setup-agents.test.ts`
- [ ] **Step 1: Run the setup-agents test file**
Run:
```bash
pnpm --filter @ktx/cli exec vitest run src/setup-agents.test.ts
```
Expected: PASS.
- [ ] **Step 2: Run the CLI type-check**
Run:
```bash
pnpm --filter @ktx/cli run type-check
```
Expected: PASS.
- [ ] **Step 3: Run the CLI test script**
Run:
```bash
pnpm --filter @ktx/cli run test
```
Expected: PASS.
- [ ] **Step 4: Run dead-code checks for TypeScript changes**
Run:
```bash
pnpm run dead-code
```
Expected: PASS. If Knip reports unrelated existing findings, inspect them and
record the exact unrelated findings before finishing.
- [ ] **Step 5: Run pre-commit for modified TypeScript files**
Run:
```bash
uv run pre-commit run --files packages/cli/src/setup-agents.ts packages/cli/src/setup-agents.test.ts
```
Expected: PASS. If the repository has no usable pre-commit environment or the
configured tool versions are unavailable, state the exact failure and keep the
passing pnpm checks as the closest available verification.
- [ ] **Step 6: Inspect the final diff**
Run:
```bash
git diff -- packages/cli/src/setup-agents.ts packages/cli/src/setup-agents.test.ts
```
Expected: The diff only contains output-polish changes on top of the existing
Claude Desktop split-ZIP baseline. It must not revert the current
`ktx-analytics.zip` and `ktx.zip` behavior.
## Self-review checklist
- The agents multiselect message is exactly
`Which agent targets should KTX install?`.
- The short navigation hint appears once for interactive no-target runs and not
for scripted target runs.
- The shared `withMultiselectNavigation` helper and its tests are unchanged.
- `Agent integration complete`, `KTX project`, and `Installed agents` no longer
appear in setup-agents install-summary output.
- `Required before using agents` remains a `note()` box.
- `All set.` appears only when the next-actions note is printed.
- Current uncommitted README and docs-site split-ZIP edits are untouched.