mattpocock /
skills
Small, editable agent skills for planning, testing and reviewing code with AI
- 301k stars
- 26.9k stars since 3 Oct
- Shell
- Updated yesterday
- Some experience
- Agent skills & plugins
Coding agents often build the wrong thing, talk too much, write code that doesn't run, or pile up messy design. This repo is the author's answer: a set of skill files, each one a short instruction set the agent follows when you type a command like /grill-me or /tdd.
The skills tackle those four problems in turn. Grilling skills have the agent question you before it builds, and keep a shared glossary and decision records so it uses your project's terms. Others push a red-green-refactor test loop, a gated bug-diagnosis routine, and regular reviews of code design.
The README contrasts it with GSD, BMAD and Spec-Kit, which it says own the whole process. These skills are meant to be small and composable, to work with any model, and to be edited. Some can only be invoked by you; others the agent can reach for on its own.
What it does
Grilling sessions
/grill-me and /grill-with-docs make the agent ask detailed questions before building. The docs version also updates a glossary and architecture decision records as you go.
Test-driven development
/tdd guides the agent through red-green-refactor, one vertical slice at a time, with guidance on good and bad tests.
Bug diagnosis loop
/diagnosing-bugs builds a failing feedback loop, then minimises, hypothesises, instruments, fixes and adds a regression test.
Spec to tickets to code
/to-spec, /to-tickets, /implement and /implement-spec turn a conversation into a spec and tickets on GitHub, Linear or local files, then build them.
Architecture and review
/improve-codebase-architecture surveys a codebase for design improvements, and /code-review checks a diff against your standards and your spec using parallel sub-agents.
What you could build
- Align with a coding agent on a feature by letting it grill you before any code is written.
- Build a feature test-first by running the /tdd skill in an existing project.
- Track down a hard bug or performance regression with a gated diagnosis loop.
- Break a large plan into tickets with blocking links on GitHub, Linear or local files.
- Review a diff against your repo's coding standards and the original spec.
Get started
Install as a Claude Code plugin for a managed bundle, or use the npx installer to copy editable skill files into your repo. Then run /setup-matt-pocock-skills once per repo to choose your issue tracker, triage labels and docs location.
claude plugins install mattpocock-skillsnpx skills@latest add mattpocock/skillsCommands are copied from the project's README. Read it on GitHub for the full setup.
Anything better than skills?
mattpocock/skills is the strongest pick for small, editable skills covering grilling, TDD, diagnosis and review. Look elsewhere if you want a full spec-to-ship lifecycle or a bundled harness with hooks and agents.
-
agent-skillsAlternative
Choose it if you want a structured lifecycle (spec, plan, build, test, review, ship) with 25 skills and slash commands that enforce small verified steps.
Trade-off: It is more prescriptive and larger, so it shapes your process more and is harder to trim to your own style.
-
ECCAlternative
Choose it if you want a ready-made, batteries-included workflow with agents, hooks, rules and memory instead of assembling a few skills yourself.
Trade-off: It is a large framework (hundreds of skills) that owns much of your process and is heavy to understand or edit.
-
skillsUse alongside
Use it alongside to install, update and remove these skills across Claude Code, Codex, Cursor and other agents with one CLI.
Trade-off: It only manages installation and adds a tool dependency; it provides no workflow skills itself.
-
improveUse alongside
Pairs well when you want a strong model to audit the repo and write plans that a cheaper model or person then implements.
Trade-off: It is narrower, covering audit and planning only, and its last commit was about six weeks ago.
More in Agent skills & plugins
superpowers
A set of skills that makes coding agents plan, test and review before they write code
ECC
Agents, skills, hooks and memory that add an engineering workflow to coding agents
andrej-karpathy-skills
A CLAUDE.md file that tells Claude Code to ask first, keep code small, edit narrowly
anthropics/skills
Anthropic's example Agent Skills: folders of instructions Claude loads per task
ponytail
A skill that makes AI coding agents write the least code that works
gstack
Slash-command skills that give Claude Code roles like reviewer, QA and release engineer
addyosmani/agent-skills
A pack of 25 Markdown skills that guide AI coding agents from spec to ship
taste-skill
Agent skills that steer AI coding tools toward better layout, type and motion
archify
Agent skill that turns a description or a codebase into an interactive HTML diagram
impeccable
A design skill for AI coding agents that steers them away from generic frontend looks