mattpocock/skills

Skills – Plug‑in Library for AI Coding Agents

What it is – A collection of skills (small, reusable command‑like scripts) that can be installed into AI‑powered coding assistants such as Claude Code, Codex, GitHub Copilot, Gemini, or any custom agent. Each skill lives behind a slash command (e.g. /grill-me, /tdd, /improve-codebase-architecture) and encodes a disciplined workflow that the agent can run automatically or the user can invoke.

Why it matters – Modern code‑generation agents are powerful but often mis‑aligned, overly verbose, or produce buggy, hard‑to‑maintain code. Matt Pocock’s skills inject classic software‑engineering practices (spec‑writing, test‑driven development, domain‑model glossaries, architecture reviews) directly into the agent’s reasoning loop, giving the model clear, token‑efficient guidance and a feedback loop that mirrors human engineering processes.

How you use it

1. Install the skill set for your chosen agent via a one‑liner (e.g. claude plugin install mattpocock-skills@claude-plugins-official). The installer copies editable files into your repository.
2. Run the setup skill /setup-matt-pocock-skills once per repo – it asks about your issue tracker, label conventions, and where to store generated docs.
3. Invoke the desired skill in the agent chat, e.g. /grill-me to interview yourself about a change, /tdd to start a red‑green‑refactor cycle, or /improve-codebase-architecture to get an HTML report of refactoring opportunities.

Key skill categories

• Engineering (user‑invoked) – workflow orchestration: grill-with-docs, triage, to-spec, implement, wayfinder, etc.
• Engineering (model‑invoked) – reusable sub‑routines the agent can call automatically: tdd, diagnosing-bugs, code-review, prototype.
• Productivity (user‑invoked) – general work‑flow helpers: grill-me, handoff, teach, wait-what.
• Productivity (model‑invoked) – primitives that power the user‑invoked commands, such as grilling and writing-for-agents.

Who it’s for – Developers who already use AI coding assistants and want a repeatable, disciplined way to keep the output maintainable, well‑documented, and aligned with project conventions.

Where to learn more – The repository’s README explains installation, the philosophy behind each skill, and links to the individual SKILL.md files that describe the exact behaviour.