A fast, searchable cheat sheet for every Claude Code surface — slash commands, CLI, keyboard shortcuts, hooks, MCP, agents, skills, settings.json keys, and environment variables. One page. Instant search. Always current.
Not affiliated with Anthropic. Community-maintained.
https://<your-username>.github.io/claude-cheat-code/
No build step. Just serve the static files:
python3 -m http.server 8080
# open http://localhost:8080- Vanilla HTML/CSS/JS, no bundler
- Tailwind CSS via Play CDN
- Fuse.js (CDN) for fuzzy search
- Data lives in
data/commands.json, validated bydata/schema.json
The page ships with full SEO metadata (Open Graph, Twitter Card, canonical URL, JSON-LD WebSite + SoftwareApplication, theme-color, social card) and a <noscript> fallback that points at plain-text directories.
For LLMs and crawlers that prefer plain text, scripts/generate-static.mjs emits these on every deploy from commands.json:
/llms.txt— short directory following the llmstxt.org convention/llms-full.txt— the entire reference as plain markdown/sitemap.xml— homepage + one URL per category hash/robots.txt— open crawl + sitemap pointer/og.svg— 1200×630 social card
These are gitignored; they only exist in the deployed site. Regenerate locally with node scripts/generate-static.mjs.
- Push the repo to GitHub. Recommended slug:
claude-cheat-code. - Settings → Pages → Source: GitHub Actions.
- Push to
main. TheDeploy to GitHub Pagesworkflow publishes the site.
The Refresh data workflow runs on the 1st of every month. It opens a stale-refresh issue if no chore: monthly refresh PR has landed in 35 days. The PR itself is opened by a remote agent configured via /schedule using scripts/refresh.md as its prompt.
See CONTRIBUTING.md. Adding a command is one JSON entry and a PR.
MIT — see LICENSE.