skills: Essential AI Agent Skills for Real Engineering
Summary
skills by Matt Pocock offers a collection of AI agent skills designed to enhance real-world engineering workflows. These skills address common pitfalls in AI-assisted development, promoting better alignment, conciseness, and code quality. Engineers can leverage these tools to build more robust and maintainable applications with their coding agents.
Repository Info
Tags
Click on any tag to explore related repositories
Introduction
mattpocock/skills is a comprehensive GitHub repository curated by Matt Pocock, offering a powerful collection of AI agent skills. Designed for "Real Engineers," these skills aim to elevate the quality and efficiency of AI-assisted software development. They tackle common challenges faced when working with coding agents, ensuring better alignment, clearer communication, and more robust code. This repository is built upon decades of engineering experience, condensed into practical, composable tools.
Installation
Getting started with mattpocock/skills is quick and straightforward, taking only about 30 seconds.
- Run the skills.sh installer:
npx skills@latest add mattpocock/skills - Pick the skills you want and which coding agents you want to install them on. Make sure you select
/setup-matt-pocock-skills. - Run
/setup-matt-pocock-skillsin your agent. It will ask you about your issue tracker, triage labels, and where to save documentation. - You are now ready to leverage these powerful skills in your development workflow.
Examples
The repository provides a rich set of skills categorized into Engineering, Productivity, and Misc. Here are some highlights demonstrating their practical application:
/grill-meand/grill-with-docs: These are crucial for ensuring alignment between you and your agent. They facilitate detailed questioning sessions, helping to clarify requirements and prevent misunderstandings before coding begins./grill-with-docsalso aids in building a shared language and documenting decisions.- Shared Language via
CONTEXT.md: A core concept embedded in skills like/grill-with-docsis the creation of a shared language. This document helps agents decode project jargon, leading to more concise communication, consistent naming, and reduced token usage. For example, replacing "There's a problem when a lesson inside a section of a course is made 'real'" with "There's a problem with the materialization cascade" significantly improves clarity and concision. /tddand/diagnose: To combat non-working code, the/tddskill promotes a red-green-refactor loop, guiding the agent to write failing tests first and then fix them. The/diagnoseskill provides a disciplined loop for tackling hard bugs and performance regressions, following steps like reproduce, minimize, hypothesize, instrument, fix, and regression-test./improve-codebase-architecture: Addressing the challenge of "ball of mud" codebases, this skill helps identify opportunities for architectural improvement, ensuring that your AI-generated code remains maintainable and well-designed. It's recommended to run this periodically to keep the codebase healthy.- Other notable skills include
/to-issuesfor breaking down plans into GitHub issues,/to-prdfor generating Product Requirement Documents, and/zoom-outfor gaining broader context on code.
Why Use
mattpocock/skills exists to address fundamental failure modes in AI-powered development. By integrating these skills, engineers can:
- Improve Alignment: Bridge the communication gap with AI agents through structured grilling sessions, ensuring the agent truly understands the desired outcome.
- Enhance Conciseness: Develop a shared language to reduce verbosity, foster consistent understanding, and optimize token usage.
- Ensure Code Quality: Implement robust feedback loops, including Test-Driven Development (TDD) and disciplined debugging, to consistently produce working and reliable code.
- Maintain Architecture: Proactively manage software entropy and invest in codebase design, preventing complexity from spiraling out of control and ensuring long-term maintainability.
These skills distill decades of engineering wisdom into actionable, repeatable practices, empowering developers to ship higher-quality applications with their AI assistants.