Skills for Real Engineers

mattpocock/skillsReviewed by AI

What it does

Matt Pocock's personal set of agent skills: about 30 short instruction files that add commands to an AI coding agent. The headline ones interview you about what you actually want (/grill-me, /grill-with-docs) before any code, then turn the conversation into a spec and tickets, build it with test-first loops, review the diff and write the pull request. Others cover debugging, codebase design and handing a session over. It is for programmers who already use Claude Code, Codex, Copilot, Gemini CLI, Cursor or a similar agent. Installing is one or two commands inside the agent.

How easy it is to install

C · install 1 or 2 tools · Windows

Windows C
macOS C
Linux C

You need a supported coding agent first, then one or two commands, for example `claude plugin install mattpocock-skills@claude-plugins-official`, or `npx skills@latest add mattpocock/skills -a <agent>` for other agents, which needs Node.js. After that you run `/setup-matt-pocock-skills` once per project and answer questions about your issue tracker, labels and where documents go. The README warns not to install both the plugin and the skills.sh copies, or you get every skill twice.

Numbers on GitHub

Stars: 308,394on Oct 31, 2026
Downloads: —all releases
Growth: +5,097stars in 6 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Other tools, Node.js
To build from source: Node.js
Disk space: 100 MB–500 MB · estimate
Accounts or keys: None
Cost: Free, API costs extra · The skills are free (MIT). They only do something inside an AI coding agent, which usually needs a paid plan or paid API use. The author also runs a newsletter and a site at aihero.dev.
License: MIT (Permissive)
Last update: Oct 31, 2026

Risk signals (5)

Attentionunusual star rate
306638 stars in 268 days (~1144 per day)
automatic check
Attentionagent writes and commits code on its own
The reference says /implement "drives /tdd at pre-agreed seams and closing out with /code-review before committing", and /implement-spec "implements a whole spec on one integration branch... running implementer subagents across the ready frontier for maximum concurrency". Several agents therefore change files and make commits in your repository in one run.
AI review
Infowrites into your issue tracker and docs
The setup skill asks "which issue tracker you want to use (GitHub, GitLab, local files, or anything else you describe)" and which labels you use; /to-spec then "publishes it to the issue tracker" and /to-tickets writes tickets "as native blocking links on a real tracker". Skills such as /grill-with-docs and /domain-modeling also edit GLOSSARY.md and architecture decision records in your repository.
AI review
Infogenerates shell scripts you then run
The /wizard skill "generates an interactive bash wizard that walks a human through steps only they can perform: provisioning infrastructure, setting up credentials or CI secrets... or running a one-off migration or cutover". You are the one who runs that generated script, so it is worth reading before you do.
AI review
Infoplugin updates itself
README Installation: the Claude Code plugin "Updates itself by default", Codex "Updates itself at startup" and the Copilot marketplace entry is configured with "autoUpdate": true. New instructions can reach your agent without you reviewing them; the skills.sh route copies editable files that you update by hand instead.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether the skills actually improve the result on a real project
How much extra AI usage, and therefore cost, the long interview and subagent flows consume
Which skills each agent other than Claude Code really picks up, and how they behave on Windows
What the automatic updates bring in later, since Claude Code, Codex and Copilot refresh the plugin by themselves
Whether the automatic code review and ticket writing match your team's process