Agent Skills

An opinionated workflow for coding agents — my own way of working, current best practices, and lessons from real projects — kept only where evals show it helps.

These skills encode how I build software with agents: scope the task, build against checks, get a second pair of eyes on the diff, prove the UI works, and keep CI green. Each one is run with and without itself; a skill that doesn’t change the outcome gets cut. They install into Claude Code, Cursor, Codex, Gemini CLI, Copilot, Windsurf, OpenCode, and any other Agent Skills-compatible tool.

git clone https://github.com/mthines/agent-skills.git
cd agent-skills && bash scripts/sync-symlinks.sh

What you get

🚀 Ship a feature from one sentence

aw scopes the job, writes a short brief with sealed checks, builds until they pass, gets a fresh-eyes review, and opens a draft PR.

Add dark mode, all the way to a PR

aw · autonomous-workflow

🐛 Fix a bug from whatever you have

A stack trace, a span URL, a screen recording, a ticket, or a hunch: it reproduces the bug, traces the root cause, fixes it behind a failing test, and opens a draft PR.

/fix-bug 

fix-bug · video-analyser

🔍 Reviews that catch real bugs

Independent finders flag, a verifier filters, and only grounded findings reach your PR. review-loop then fixes or answers every thread until the PR converges; with no PR yet, it reviews the local branch.

/pr-review 42
/review-loop

pr-reviewer · review-loop

✅ Proof the UI actually works

The brief’s spec runs on your local stack or the preview and tries to break the change. Then the telemetry is read back to check that every click reached the API and the database the way it should. Screenshots, recordings, and the reasoning behind each check land in one PR comment. Run setup once per repo; it commits .agents/aw/.

/aw-verify setup
/aw-verify preview 42

aw-verify · ui-verify · telemetry-verify