mattpocock/skills
A pack of reusable prompt playbooks that keep AI coding agents like Claude Code from going off the rails
mattpocock/skills is a collection of 'skill' files that guide AI coding agents (Claude Code, Codex, etc.) to interview you before coding, share consistent project terminology, run test-driven development loops, and periodically review codebase architecture. Each skill is a small, editable instruction document rather than a rigid all-in-one framework. It can be installed as a managed Claude Code plugin or copied into a project as editable files via the skills.sh installer.
What it does
- Targets four common failure modes with AI coding agents: misunderstanding what you want, overly verbose responses, code that doesn't actually work, and codebases that turn into unmanageable messes over time.
- Skills like /grill-me and /grill-with-docs make the agent interrogate you about your plan before writing code, and /grill-with-docs also builds a shared CONTEXT.md vocabulary and ADR (architecture decision record) documents so agent responses get shorter and more precise over time.
- The /tdd skill enforces a red-green-refactor loop (write a failing test, then fix it) so the agent gets consistent feedback instead of guessing, while /diagnosing-bugs applies a disciplined step-by-step debugging process.
- /improve-codebase-architecture periodically scans a codebase for places where the design could be simplified ('deepening opportunities') and presents them as an HTML report, though the author notes it's a survey tool, not an automatic fixer for old, tangled code.
- Skills are split into 'user-invoked' ones you type explicitly (like /triage or /to-spec) and 'model-invoked' ones the agent can call automatically when relevant (like tdd or code-review), and installation is available via Claude Code plugin (auto-updating, read-only) or the skills.sh CLI (editable, forkable files).
Why it matters
As AI coding agents become a bigger part of everyday software work, having a lightweight, model-agnostic set of practices to keep them aligned, disciplined, and architecture-aware matters more than another rigid AI framework that takes control away from the developer. With 224,352 stars, this repo is being widely adopted as a practical reference for how experienced engineers actually structure their AI-agent workflows.
Terms in this repo
- Claude Code plugin · Anthropic의 코딩 에이전트 Claude Code에 통째로 설치되는, 자동 업데이트되는 읽기 전용 확장 기능 묶음
- skills.sh · skill 파일들을 프로젝트에 직접 복사해 넣어 사용자가 편집할 수 있게 해주는 설치 도구
- red-green-refactor · 실패하는 테스트를 먼저 작성한 뒤(red) 통과시키고(green) 코드를 정리하는(refactor) 테스트 주도 개발 절차
- ADR (Architecture Decision Record) · 설계 관련 결정과 그 이유를 기록해두는 문서 형식
- CONTEXT.md · 프로젝트에서 쓰는 전문 용어와 도메인 개념을 정리해 에이전트와 사람이 같은 언어로 소통하게 돕는 문서
Repository description (English)
Skills for Real Engineers. Straight from my .agents directory.
Open on GitHubTrending repos
- vorssaint/vorssaint-utilsOne free menu bar app replaces a dozen paid Mac utilities
- Alishahryar1/free-claude-codeA local proxy that lets coding AI agents run on 49 free or cheap model providers instead of one paid service
- freestylefly/awesome-gpt-image-2A library of 532 reverse-engineered prompts that turn GPT-Image2 into a predictable image-making tool
- block/buzzAn open-source workspace where humans and AI agents chat, code, and review in the same rooms
- NousResearch/hermes-agentNous Research's Hermes is an AI agent that gets smarter the more you use it
- virgiliojr94/book-to-skillA tool that turns technical book PDFs into on-demand reference skills for AI coding agents
- VoltAgent/awesome-agent-skillsA single hub collecting over 1000 'how-to' manuals that make AI coding assistants act like experts
- anthropics/claude-plugins-communityA shared shelf where anyone's Claude add-ons get listed for install