Agentpedia Codes
Directory of installable AI agent skills for claude, codex
agentskills.codes is an open registry for AI coding agent skills — reusable SKILL.md instruction packs that extend what Cursor, Claude Code, Codex, and GitHub Copilot can do. We index 19,000+ skills discovered from GitHub, enrich each with categories, use cases, safety signals, and live source content, and let developers browse, compare, and install with a single command. No account required — search the catalog, open a detail page, copy a zip install or npx degit command, and drop the skill into your project or global agent config in minutes.
Features
- 19,000+ skill catalog — continuously updated from GitHub SKILL.md discovery
- Multi-client install — Cursor, Claude Code, Codex, and GitHub Copilot (project + global scope)
- One-command install — hosted zip download via API or degit from source repo
- Rich skill pages — TL;DR, what it does, use cases, trigger phrases, tags, stack/tool facets
- Trust & safety panel — capability flags, clean/review/warn grades, source health (ok/moved/gone)
- Source transparency — GitHub stars, license, last commit date, token footprint estimate
- Live content — fetches the real SKILL.md (or README) from GitHub at request time
- Faceted browse — categories, tags, “for tool,” stack pages, and leaderboard
- Search API — public /api/v1/skills with query, sort, category, and pagination
- Programmatic SEO — sitemap, JSON-LD, canonical URLs for discoverability
- Daily auto-discovery — GitHub Actions pipeline finds and enriches new skills automatically
- Open source — MIT licensed, no auth, no database — catalog is the system of record
Use Cases
- Find a skill fast — search “debug Solana,” “write blog post,” or “security audit” and get ranked results with install commands
- Compare before you install — read use cases, trigger phrases, and safety grades on one page instead of cloning random repos
- Onboard a new teammate — share a skill URL; they install in one command instead of hunting GitHub
- Standardize team workflows — pick vetted skills for code review, testing, docs, or deployment and install project-wide
- Explore by stack or tool — browse skills tagged for Next.js, Rust, Docker, Linear, Sentry, etc.
- Discover trending skills — use the leaderboard to see what’s popular by installs and views
- Install dead or moved repos safely — hosted zip still works when upstream is gone; trust panel flags source health
- Build on the API — integrate skill search into your own dashboard, IDE plugin, or internal tooling
- Publish your skill — if your repo has a SKILL.md in a standard path, daily discovery can list it automatically
- Evaluate agent capabilities — see what skills exist for a category (testing, DevOps, design, crypto) before committing to a workflow
Comments
The trust panel is the part I'd most want to understand. A SKILL.md is instruction text an agent will actually execute, so an auto-discovered catalog of 19,000 of them is a real prompt-injection surface. How is the clean/review/warn grade derived — static heuristics over the file, a model pass, or human review? And is a skill re-graded when the upstream repo changes, or is the grade pinned to the commit you indexed? That second one matters more because of the hosted-zip fallback: the copy you serve can outlive both the review and the repo it came from.
Agentpedia Codes Your product has strong potential, but I found a few key improvements that could make it even better. I'd love to share my feedback and suggestions—please contact me at [email protected]
Premium Products
Sponsors
Buy
Makers
Makers
Comments
The trust panel is the part I'd most want to understand. A SKILL.md is instruction text an agent will actually execute, so an auto-discovered catalog of 19,000 of them is a real prompt-injection surface. How is the clean/review/warn grade derived — static heuristics over the file, a model pass, or human review? And is a skill re-graded when the upstream repo changes, or is the grade pinned to the commit you indexed? That second one matters more because of the hosted-zip fallback: the copy you serve can outlive both the review and the repo it came from.
Agentpedia Codes Your product has strong potential, but I found a few key improvements that could make it even better. I'd love to share my feedback and suggestions—please contact me at [email protected]
Premium Products
New to Fazier?
Find your next favorite product or submit your own. Made by @FalakDigital.
Copyright ©2025. All Rights Reserved