AI-agent skills

Browse published agent skills for Claude Code, Codex and other clients. Compare sources, licenses and setup instructions, or draft your own SKILL.md.

Agent worktree isolation

Run each coding agent in a separate git worktree to avoid dirty-checkout conflicts and accidental overwrite of another run.

Source: Git worktree docs · License: GPL-2.0 / Git docs

Review pull requests with grounded diffs, security checks, inline comments, and CI-aware follow-up tasks.

Source: GitHub CLI + repository diff · License: MIT / project license

Drive a real browser for UI smoke tests, console/network inspection, screenshots, and regression evidence.

Source: Playwright docs · License: Apache-2.0

Reproduce bugs with focused tests, fix root causes, and keep verification commands attached to the report.

Source: Node.js test runner docs · License: MIT / Node.js docs license

Expose read-only product data to agents through MCP tools with schemas, resources, and no invented results.

Source: Model Context Protocol docs · License: Protocol docs / implementation license

Guide agents through bring-your-own-key flows with masking, no echo, revoke paths, and encrypted-at-rest storage requirements.

Source: OWASP guidance · License: CC BY-SA 4.0