Skip to content
← Tools / Skills
Skill · .md packageMatt Pocock/10 trustVerified
Skills for Real Engineers logo

Skills for Real Engineers

Matt Pocock's battle-tested agent skills — TDD, specs, debugging, from a working engineer's setup.

Visit official site →

About

Skills for Real Engineers is Matt Pocock's (Total TypeScript, AI Hero) public collection of agent skills, pulled straight from his own .agents directory. The philosophy is explicit: real engineering, not vibe coding — small, composable, model-agnostic skills based on decades of engineering practice (tracer bullets, TDD, ubiquitous language, deep modules). They target the common failure modes he sees in Claude Code, Codex, and other agents: agents not doing what's asked, excessive verbosity, broken code, and balls of mud. The repo is one of the most-starred agent-skill collections on GitHub (~240k stars), MIT-licensed, installable as a Claude Code plugin, with ~20M installs via skills.sh. A setup skill walks you through choosing an issue tracker, labels, and docs locations.

The honest picture: these are opinionated workflows from one engineer's practice — they impose process (specs, tickets, TDD) that can feel heavy for quick scripts or explorations. The skills assume you buy into the methodology; cherry-picking individual skills works but loses some coherence. As with any popular repo, verify you are installing from the genuine mattpocock/skills source, since forks and copycats exist.

What it can do

  • Spec-driven development skills
  • TDD and debugging workflows
  • Issue triage and ticket management
  • Agent verbosity and quality guardrails
  • Cross-agent installation via skills.sh

Setup

Run npx skills@latest add mattpocock/skills, pick skills and target agents, then run /setup-matt-pocock-skills in your agent to configure tracker and labels.

Prerequisites

A supported coding agent (Claude Code, Cursor, Codex, etc.); willingness to adopt ticket/spec workflows.

Use cases

coding-agentstddspec-driven-devdebugging

Works with

Trust score — 9/10

/10 trustVerified

Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 50; Trending score 92/100 from researcher assessment.

★ 241,800 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.