Project Handbook — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Project Handbook (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
Aggregate score unchanged between these scans.
The primary manifest — the file an agent reads to learn what this artifact does.
A Claude Code skill that turns any JS/TS project into a dual-audience handbook: a browsable HTML site for humans, backed by markdown/JSON data files that AI agents read directly. One set of files, two audiences, never out of sync.
▶ [Live Demo](https://gexianss.github.io/project-handbook/) — browse a generated handbook right in your browser, no install needed.
Demo: overview, architecture diagrams, dev guide, health dashboard
freshness & vulnerabilities, git hotspots, testing gaps. Real tools first (npm audit, knip, git log); honest unknown when a tool can't run.
the backend, how the team works, machine capacity. Questions are generated from the scan, asked in one batch. Unknown answers become a ⚠️ To-confirm list with pointers on where to find out, not guesses.
index.html + viewer — sidebar navigation, Mermaid architecture diagrams, healthdashboard. Fully offline (libraries vendored).
data/*.md + health.json — the actual content. AI agents read these directly;the viewer renders the same files for humans.
open-handbook.bat / .sh — double-click to serve and open (needs Node ≥ 18).The viewer template never changes after generation. AI maintains only data/ — so updating the handbook is editing a few markdown files, and the site is instantly current.
Every page carries a freshness mark — "Updated <date> · verified at <commit>" — and re-running on an updated project does a diff-driven audit: only sections whose related files actually changed get re-read and rewritten; the rest are proven current by the git diff itself.
The viewer loads data/*.md with fetch() at runtime, and browsers block fetch on file:// pages — so the handbook must be opened over HTTP. Any of these works:
docs-site/open-handbook.bat (Windows) or rundocs-site/open-handbook.sh (macOS/Linux) — starts the bundled server and opens your browser. Needs nothing but Node.
docs-site/index.html →Open with Live Server. Bonus: the page auto-reloads whenever a data/*.md file is saved.
node docs-site/serve.cjs 8787, then openhttp://localhost:8787.
| File | Why it exists |
|---|---|
data/ | The handbook content (md/json) — the only thing AI ever edits |
index.html + assets/viewer.* | The viewer: sidebar, Mermaid rendering, health dashboard |
assets/vendor/ | marked + mermaid bundled locally (~2.6 MB) so the handbook works fully offline — no CDN, no npm install |
serve.cjs | Zero-dependency static server, because browsers block fetch on file:// |
open-handbook.bat / .sh | Double-click launchers for Windows / macOS / Linux — for readers without a dev setup |
/plugin marketplace add Gexianss/project-handbook
/plugin install project-handbook@project-handbook-marketplaceOr manually: copy skills/project-handbook/ into ~/.claude/skills/.
In any JS/TS project, ask Claude Code:
Generate a project handbook
or in Chinese: 幫專案做手冊 / 專案體檢. Re-running on a project that already has a handbook offers an audit & refresh instead of overwriting.
gracefully to AI-only analysis.
English.
examples/demo/ is a generated handbook for a fictional app — open it with examples/demo/open-handbook.bat (or .sh).
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.