AI Agent Skills
Reusable prompts and workflows for AI coding agents. Copy the content into chat or drop into your tool's skills folder. Works with Claude Code, Cursor, AMP, Windsurf, and others.
/debug
DebugAdd structured logging to track down bugs. Uses a local log file and an optional HTTP server for browser-side logging.
clicking 'Save' shows success toast but data isn't persistedWebSocket messages are received but UI doesn't update/council
ExplorationSpawn multiple agents to deeply explore a codebase area before acting.
💡 Use params like n=10 to control agent count
/council n=15, how does authentication work?find all places we use InstancedGeometry /council n=5/iterate-browser
ExplorationAutonomously iterate on UI changes using console.log and browser tools.
💡 Requires browser automation tools (MCP or built-in)
/iterate-browser the submit button doesn't close the modaldropdown menu is positioned off-screen /iterate-browser/fix-ci
PR & GitAnalyze CI failures and automatically apply fixes based on error logs.
/fix-ci/fix-ci for the latest run/implement-feedback
PR & GitAddress PR review feedback by validating each item against the diff, then implementing fixes or reporting back.
/implement-feedbackapply these PR comments/branch-description
PR & GitGenerate structured PR descriptions from git changes and copy to clipboard.
/branch-descriptionwrite a PR description for this branch/resolving-merge-conflicts
PR & GitResolve git merge conflicts by understanding branch intent and integrating with base branch.
/resolving-merge-conflictsresolve these merge conflicts/scallywag-mode
🦜Switch the whole conversation to pirate theme. All AI agents become crew members with pirate-themed names.
/scallywag-modeactivate scallywag mode/you-might-not-need-an-effect
Code qualityReview React code for unnecessary useEffect usage based on react.dev guidelines.
💡 Run with /council to find issues across the codebase
/you-might-not-need-an-effect/you-might-not-need-an-effect scope=diff to mainHow to use
Copy the content and paste directly into chat, or save as a markdown file in your project for the agent to reference.
Skills go in .cursor/skills/. Ones with a slash name (e.g. /branch-description) can also live in .cursor/commands/ and be invoked as /command-name.
Check your tool's docs for custom instructions or rules files. Most support markdown-based prompts.
