Matt Pocock · 2026-07-08 · major
mattpocock/skills 1.1.0 — engineering skill pack for AI coding agents
Matt Pocock's skill pack for AI coding agents ships 1.1.0, promoting /code-review and /wayfinder into the main suite and adding new /research and /prototype skills. The repo now sits at 164,551 stars and 14,200 forks under MIT license.
Matt Pocock's composable skill pack teaches AI coding agents how to plan, TDD, and review like a senior engineer.
Quick facts
| Maker | Matt Pocock |
|---|---|
| License | MIT |
| Latest version | 1.1.0 (Jul 8, 2026) |
| Stars | 164,551 (Jul 10) |
| Language | Shell + JavaScript |
| Agent support | Claude Code, Cursor, Copilot, Cline, Gemini CLI |
| New in 1.1.0 | /research, /prototype, /code-review graduated, /wayfinder mature |
What is it?
Version 1.1.0 of mattpocock/skills promotes /code-review and /wayfinder into the main engineering suite and adds new /research and /prototype skills for autonomous investigation and design exploration. The pack lives at github.com/mattpocock/skills, ships under MIT license, and is agent-agnostic — the README states it 'works with any model,' though it comes from Matt Pocock's own Claude directory.
How does it work?
Each skill in mattpocock/skills is a small, composable module — a shell script plus prompt files — that a coding agent invokes as a slash command. Engineering skills like /grill-with-docs, /tdd, /implement, and /code-review align the agent with the developer, run red-green-refactor loops, or apply a fixed 12-code-smell baseline. Productivity skills like /grill-me and /handoff prepare context transfer between agents. Skills chain so a spec becomes tests becomes code becomes a review.
Why does it matter?
AI coding agents often fail in four ways — misaligned with the developer, too verbose, buggy code, drifting architecture. mattpocock/skills packages the interventions a senior engineer would apply — deliberate alignment sessions, TDD, seam-based tests, a fixed smell list — into slash commands anyone can install. That turns 'get Claude to code well' from a personal prompt-engineering exercise into a shared, versioned toolkit, and the repo now sits at 164,551 stars and 14,200 forks.
Who is it for?
engineers using Claude Code, Cursor, Copilot, Cline, or other coding agents
Frequently asked questions
- Which coding agents does mattpocock/skills work with?
- mattpocock/skills is agent-agnostic — the README states it 'works with any model.' The pack was written from Matt Pocock's own Claude Code setup, but the same slash commands drop cleanly into Cursor, GitHub Copilot, Cline, Gemini CLI, and other coding agents. The skills are shell scripts plus prompt files, so anything that can invoke a slash command can run them.
- What's new in mattpocock/skills 1.1.0?
- mattpocock/skills 1.1.0 promotes /code-review from work-in-progress into the main engineering suite with a 12-smell Fowler baseline, graduates /wayfinder for decision mapping, and adds new /research and /prototype model-invoked skills. It also renames /to-prd to /to-spec and merges /to-plan and /to-issues into a single /to-tickets command.
- How is engineering different from productivity in mattpocock/skills?
- mattpocock/skills splits into two families. Engineering skills — /grill-with-docs, /tdd, /implement, /code-review, /improve-codebase-architecture — cover the software work an agent does at the keyboard. Productivity skills — /grill-me, /handoff — cover the setup around it, aligning humans with agents and passing context between agents. Both ship as user-invoked slash commands under MIT license.
Try it
git clone https://github.com/mattpocock/skills