AGENT SKILL

handbook-sync-audit

Handbook Sync Audit — verifies that generated handbooks stay aligned with their sources across three axes: (1) content reflection against the ai-workspace upstream (docs/designs, docs/adr), (2) structural linkage within and between handbook repositories, and (3) section-level language parity (ko/en/ja/es) that pinpoints exactly which sections a translation is missing. Responds to "audit handbook", "handbook parity check", "handbook sync audit", "textbook drift check".

License 许可
AGPL-3.0
最近更新
2026-11-14
Stars
★ 25
Forks
9

01 前言 Front Matter

name
handbook-sync-audit
owner
pm
scope
common
status
active
version
1.0.5
metadata
{ "triggers": [ "audit handbook", "handbook parity check", "handbook sync audit", "textbook drift check" ] }
description
Handbook Sync Audit — verifies that generated handbooks stay aligned with their sources across three axes: (1) content reflection against the ai-workspace upstream (docs/designs, docs/adr), (2) structural linkage within and between handbook repositories, and (3) section-level language parity (ko/en/ja/es) that pinpoints exactly which sections a translation is missing. Responds to "audit handbook", "handbook parity check", "handbook sync audit", "textbook drift check".
l2_propagate
true
last_reviewed
1787961600000
prerequisites
handbook (audits the output of the H-Stage handbook pipeline)

03 SKILL.md 正文

Context

Handbooks drift in three predictable ways: the upstream workspace gains new concepts the handbook never mentions, handbook pages get restructured and links break (including links between sibling handbooks), and new sections are added to a base-language page without being translated. This skill runs one self-contained audit script that detects all three and reports findings with enough detail (missing section ids, unreflected source docs, broken link targets) to investigate the cause and plan the fix.

Works on any handbook directory with the co-deck layout (<repo>/docs/**/*.html, language variants suffixed _en/_ja/_es) — both freshly generated co-deck handbooks and existing handbook repositories such as Handbooks/multi-agent-harness-handbook.

When to Use

  • After generating or updating a handbook (post H-Stage verification)
  • Periodic drift check against the workspace (recommended: monthly)
  • User says "audit handbook", "handbook parity", "textbook drift check"

Execution Steps

Run the audit script via bun (per ADR-0036):

# 1. Content reflection — does the handbook mention upstream workspace topics?
bun scripts/handbook/handbook-sync-audit.ts content \
  --handbook-dir <handbook>/docs \
  --workspace <ai-workspace-root> \
  [--since 2026-08-01]

# 2. Structural linkage — broken links, orphan pages, cross-handbook links
bun scripts/handbook/handbook-sync-audit.ts structure \
  --handbook-root <dir-containing-handbook-repos>

# 3. Language parity — which h2/h3 sections each translation is missing
bun scripts/handbook/handbook-sync-audit.ts parity \
  --handbook-dir <handbook>/docs

# Or all three at once (exit 1 if any FAIL finding):
bun scripts/handbook/handbook-sync-audit.ts all \
  --handbook-dir <handbook>/docs --workspace <ws> --handbook-root <root>

Interpreting Results

Mode FAIL means Typical cause Typical fix
content Workspace design docs / ADRs whose keywords the handbook never mentions Upstream feature landed after the handbook was written Dispatch handbook-writer (or the variant's authoring specialist) to add/refresh the relevant chapter section
structure Broken internal links, orphan pages, dead cross-handbook links Pages renamed/moved without link updates; new page never linked from index Fix hrefs; link the orphan from index.html/nav
parity Sections present in the base page but missing from _en/_ja/_es Content added to base page only Translate the missing section into each variant, mirroring heading ids

Finding → Fix Loop

  1. Run the audit and capture the findings list.
  2. For each finding, investigate the cause (when was the source added? was the page moved? is the translation pending?) and record it in the fix commit.
  3. Apply fixes (docs-writer / handbook-writer territory), then re-run until the audit reports PASS.

Output Format

The script prints a per-mode report and exits 0 (PASS) or 1 (findings exist):

═══ <mode> audit — <description> ═══
handbook docs: <path>
...per-item findings (missing sections with #ids, unreflected source docs,
broken link targets)...
result: PASS | FAIL (findings above — investigate causes, then resolve)

When dispatching handbook-writer after an audit, include the captured findings block verbatim — the #section-id and source-doc references are the fix plan.

Related Skills

  • handbook — H-Stage generation pipeline this skill audits
  • project-review — broader multi-agent project review that can include handbook output
  • markdown link validation for workspace docs runs through the handbook doctor pipeline, not a separate skill

Notes

  • The content mode uses keyword-quorum matching (≥2 slug keywords of a source doc must appear in the handbook text) — it is a drift detector, not a semantic proof. Review its misses before acting: a keyword hit on a generic term can be a false negative trigger for investigation.
  • parity compares heading ids; renaming a section id counts as a parity finding for the other languages. Keep ids stable across languages.
  • Exit code 1 = findings exist; CI can wire this script as a scheduled gate.

04 同仓库更多 Skills

5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25