AI Agent Skills for
521+ production-grade, hand-crafted procedures — from RAG pipelines and incident response to GxP documentation, contract triage, and lesson planning. Give your agent the working knowledge of a whole senior team.
Works with OpenCode · Claude Code · Cursor · Codex · Copilot · Gemini CLI · Windsurf & more
How your agent thinks with skills
Progressive disclosure: skills cost zero context until the exact moment they matter.
Discovery
At startup your agent loads only each skill's name + description — a few tokens per skill.
Activation
You ask "design our chunking strategy" — the agent matches intent and loads the full procedure.
Execution
Steps, pitfalls, commands, and scripts execute in context. Your agent works like the senior who wrote it.
Browse by profession
47 tracks spanning software, science, medicine, law, business, and beyond — every one organized into concrete domains.
Signature skills
Authored directly from real production workflows.
Choose chunk sizes, overlaps, and structure-aware splits so retrieved context actually helps the model answer.
Accept retried mutations gracefully by keying operations so duplicates return the original result.
Authenticate Vercel CLI via a stored token instead of browser OAuth — the reliable path when clock drift breaks login flows. Use when deploying, linking project
A merge-rate-first contribution workflow: issue selection, minimal diffs, review follow-through. Use when contributing to someone else's repo, opening PRs upstr
Verify claims, detect namesake contamination, and flag privacy leaks across public profiles. Use during background reviews, bio writing, personal-brand cleanups
Structure bug-hunt experiences into engaging technical posts that teach method, not just drama. Use when writing dev blogs, hackathon retrospectives, or 'how I
Skills beat copy-pasted prompts
Zero context tax
Prompts burn tokens every turn. Skills load once when relevant — run dozens without bloating context.
Versioned & shareable
Git-backed markdown. Review skill changes like code, ship improvements to your whole team at once.
Agent-portable
One SKILL.md works across 75+ agents via the open format. No vendor lock-in on your knowledge.
Procedural, not vibes
Numbered steps, pitfalls, KPIs, copy-paste commands — judgment encoded, not motivational quotes.
Fresh off the bench
Latest tracks added to the library — expansion packs ship continuously.
Give your agent a brain upgrade
Free, MIT-licensed, built in the open. Star the repo, install a skill, or contribute your own playbook.