[Feature]: Add /api/skills endpoint + browsable web UI for the skills registry

[justelson]

---

Problem

Discovering skills currently requires running npx skills find <query> in a terminal or manually browsing GitHub repos. There is no structured, publicly accessible API that exposes the registry data, which makes it difficult to build UIs, integrations, or dashboards on top of it. The community is effectively blocked from composing new tooling around skills without scraping GitHub directly.

---

Proposed Solution

Add a /api/skills endpoint that returns a structured JSON list of all indexed skills, and a /browse route on skills.sh that serves as a browsable UI built on top of it.

Endpoint:

GET /api/skills?q=react&agent=claude-code

[
  {
    "name": "vercel-react-best-practices",
    "description": "React and Next.js performance optimization guidelines",
    "repo": "vercel-labs/agent-skills",
    "url": "https://skills.sh/vercel-labs/agent-skills/vercel-react-best-practices",
    "agents": ["claude-code", "cursor", "codex"],
    "tags": ["react", "nextjs", "performance"],
    "installCount": 4821
  }
]

UI (/browse): A card grid of all indexed skills with filtering by agent, tag, and keyword. Each card shows the description, install count, source repo, and a one-click copy for the npx skills add command.

---

Alternatives Considered

---

• Leaving discovery CLI-only — works today but limits the ecosystem to terminal users and blocks community tooling
• Community scraping GitHub directly — fragile, puts unnecessary load on the GitHub API, and duplicates logic already inside this repo
• A separate community-built registry — possible but would fragment the ecosystem; better to have one canonical source

---

Additional Context

Happy to open a PR for this. Before doing so, a few questions for maintainers:

1. Where does the registry index currently live — flat file, database, or GitHub API at runtime?
2. Is the skills.sh frontend source part of this repo or a separate private deployment?
3. Any preferred shape or constraints for the API response?

---

Note I checked and;

From the README:

The only current discovery mechanism is npx skills find [query] — a CLI-only, interactive search. There is no HTTP equivalent. The skills.sh site is referenced as the canonical directory but exposes no public API. The proposed endpoint would be a direct complement to skills find, making the same data accessible over HTTP. The API should also respect the metadata.internal flag — hiding internal skills by default, consistent with existing CLI behavior.