AI Agent Skills Catalogs
Table of content
skills catalogs are the emerging app store for AI coding agents. instead of fine-tuning or complex prompt chains, you drop a .md file into your project and the agent learns a new capability.
first spotted in signals — 2026-02-10 .
quick comparison
| Project | Stars | Agent | Format | Focus | Open Source |
|---|---|---|---|---|---|
| OpenAI Skills | 7.6K | Codex | .md | official skill definitions | Yes |
| Awesome Claude Skills | 33K | Claude Code | .md | curated community directory | Yes |
| Compound Engineering Plugin | 8K | Claude Code | plugin | workflow orchestration | Yes |
openai skills
the official skills catalog for OpenAI’s Codex agent. each skill is a markdown file that defines what the agent can do, how to do it, and what constraints apply.
why it matters: OpenAI chose .md files for agent customization. not JSON schemas, not YAML configs, not a GUI. plain text that humans can read and edit. this is the same bet the AGENTS.md ecosystem made — and now the largest AI company is validating it.
strengths: → official OpenAI backing → clear format and conventions → growing community contributions
weaknesses: → Codex-specific (doesn’t directly work with other agents) → relatively new, catalog is still thin
awesome claude skills
→ github.com/ComposioHQ/awesome-claude-skills
a curated list of Claude skills, resources, and tools for customizing Claude AI workflows. community-maintained by Composio.
why it matters: at 33K stars, this is the largest aggregation of agent skill patterns. it’s become a de facto directory for “what can I make Claude do?”
strengths: → massive community → covers skills, integrations, and workflow patterns → good for discovering what’s possible
weaknesses: → curation quality varies (awesome-list problem) → some entries are more promotional than practical
compound engineering plugin
→ github.com/EveryInc/compound-engineering-plugin
the official Claude Code plugin for compound engineering workflows, from Every (the newsletter/media company). enables multi-step agent workflows where outputs from one step feed into the next.
why it matters: this is the first major plugin that treats Claude Code as an orchestration platform, not just a coding assistant. the pattern — write a plugin that extends what your agent can do — is likely to become standard.
strengths: → well-designed plugin architecture → practical compound workflows → good documentation
weaknesses: → Claude Code specific → requires understanding of plugin system
the pattern
three different companies, three different approaches, one shared insight: the future of agent customization is structured skill files, not fine-tuning.
the format is .md. the distribution is GitHub. the curation is community-driven.
this is what an app store looks like before anyone builds the store.
self.md take
skills catalogs solve the “cold start” problem for AI agents. instead of spending hours writing custom prompts, you browse a catalog, drop a file in your project, and your agent gains a new capability.
the risk: skill quality. a bad skill file can make your agent worse, not better. and right now there’s no verification, no testing, no “app review.”
the opportunity: whoever builds the trusted, curated skills marketplace — with testing, ratings, and compatibility guarantees — builds the app store of the agentic era.