skills.yusuke.run

A personal, curated collection of skills — small, focused Markdown documents that describe how to do specific things with specific tools. Designed to be consumed by coding agents (Claude Code, Cursor, etc.) and by humans.

Deployed at https://skills.yusuke.run.

What is a "skill"?

A skill is a SKILL.md Markdown document that answers a focused question like:

• "How do I build a server-rendered UI with Hono?"
• "How do I combine Inertia.js with Hono?"
• "How do I structure a Cloudflare Workers project with D1?"

Each skill is small, opinionated (it reflects how I like to do it), and includes links to the canonical human-facing documentation so an agent can drill deeper when needed. The on-disk layout (skills/<name>/SKILL.md) matches the vercel-labs/skills CLI, so the catalog is also installable into Claude Code, Cursor, and other agents via npx skills add.

How agents should use this

There are two ways to consume the catalog. Pick whichever the agent / tool supports.

Fetch on demand (no installation). Drop this into your agent at the start of a new project:

fetch https://skills.yusuke.run/start.md to start a new project

The landing page (/) shows this same line, ready to copy. From /start.md the agent discovers every available skill and drills down into the ones it needs.

Install as local skills. For tools that integrate with the skills CLI (Claude Code, Cursor, OpenCode, Codex, ...):

npx skills add yusukebe/skills

This installs the catalog into the agent's local skills directory so the agent can match on description without an HTTP fetch.

URL shape

• GET / — human-facing landing with the copy-pasteable agent prompt.
• GET /start.md — agent entry point. Markdown: usage instructions + skill index.
• GET /skills/:name — a single skill (Markdown).
• GET /skills/:name.json — the same skill as JSON (frontmatter + body).
• GET /skills.json — machine-readable full index.

All Markdown content is served as text/markdown; charset=utf-8.

Repository layout

.
├── AGENTS.md                # Instructions for AI agents working in this repo
├── README.md                # This file
├── skills/                  # One directory per skill, each containing SKILL.md
│   └── <name>/SKILL.md
├── scripts/
│   └── build-registry.ts    # Scans skills/*/SKILL.md, writes src/registry.generated.ts
├── src/
│   ├── index.ts             # Hono routes
│   ├── landing.ts           # Renders the / and /start.md responses
│   ├── landing.template.md  # Template for /
│   ├── start.template.md    # Template for /start.md (with {{skills}} placeholder)
│   ├── frontmatter.ts       # Minimal YAML frontmatter parser
│   ├── registry.generated.ts # Build output — frontmatter for every skill (gitignored)
│   └── skills.ts            # Reads the generated registry, fetches bodies via env.ASSETS
└── wrangler.jsonc

Adding a new skill

See AGENTS.md for the full contribution procedure. The short version:

1. Create skills/<name>/SKILL.md with the frontmatter shown in AGENTS.md.
2. bun run dev to preview (the registry is regenerated automatically from skills/*/SKILL.md).
3. Push to main — Cloudflare Workers Builds handles the deploy.

From another working project, an agent can also open a PR for you — see the contribute skill.

Develop

bun install
bun run dev      # local preview at http://localhost:8787

Deploy

Deployment is automated via Cloudflare Workers Builds. Any push to main builds and deploys the Worker.

Initial setup (one-time, in the Cloudflare dashboard):

1. Workers & Pages → the skills Worker → Settings → Builds.
2. Connect the GitHub repository yusukebe/skills.
3. Branch: main.
4. Build command: leave empty (or bun install).
5. Deploy command: bun run deploy.

bun run deploy regenerates src/registry.generated.ts from the current skills/*/SKILL.md and then runs wrangler deploy. The build step is bundled into the deploy script so a manual local deploy can't forget it either.