TL;DR
Matt Pocock's .claude skills directory hit 74.7k GitHub stars, offering structured workflows that address the four most common ways AI coding agents fail.
Read next
Matt Pocock made his personal .claude directory public and hit 857 new GitHub stars in a day. Here's what 18 production-tested Claude Code skills look like from a prolific TypeScript educator.
5 min readThe mattpocock/skills repository hit 62k stars on GitHub's weekly trending list with 17 composable Claude Code agent skills that tackle the four failure modes killing real engineering teams.
5 min readobra/superpowers picked up 1,600+ GitHub stars in a single day by solving the real problem with coding agents - they skip the boring parts that make software actually work.
6 min readMatt Pocock built his reputation teaching TypeScript. His Total TypeScript course and newsletter have more than 60,000 subscribers. So when his GitHub repository mattpocock/skills landed at the top of GitHub Trending today with 3,886 new stars in 24 hours and 74.7k total stars overall, the description caught developers off guard: "Skills for Real Engineers. Straight from my .claude directory."
The repo is not a TypeScript utility. It is a curated set of structured workflows for AI coding agents - specifically targeting the four failure modes Pocock has identified in his own day-to-day work with Claude Code and similar tools. The premise is that agents fail predictably, and structured skills can fix that failure at the workflow level rather than the prompt level. The 6.4k forks the repo has accumulated suggest a lot of developers are personalizing it rather than just starring it.
The mattpocock/skills repository ships 18 skills divided into three categories: engineering, productivity, and miscellaneous tooling. Each addresses a named failure mode.
The engineering skills target the hardest parts of collaborative coding with an agent:
diagnose walks an agent through a formal debugging sequence - reproduce the issue, minimize the repro case, form a hypothesis, instrument the code, apply a fix, and confirm with a test. No guessing.grill-with-docs runs a structured requirements interview that writes its findings into a CONTEXT.md file and creates Architecture Decision Records (ADRs), so the context persists beyond a single session.tdd enforces a red-green-refactor loop rather than letting the agent jump straight to implementation.triage provides a state machine for labeling and categorizing GitHub issues using configurable label sets.improve-codebase-architecture uses the domain language and ADR context built by other skills to suggest architectural improvements grounded in project history.to-prd converts conversation context into a structured Product Requirements Document formatted as GitHub issues.to-issues breaks a specification into independent, vertical-slice tickets sized for parallel work.zoom-out instructs the agent to surface system-level context before diving into unfamiliar code.prototype generates throwaway prototypes - terminal apps or UI variations - for fast design validation before committing to an implementation.The productivity layer handles the overhead of long agent sessions. grill-me keeps asking clarifying questions until all decision branches resolve. caveman switches the agent to ultra-compressed output that cuts token usage by roughly 75 percent. handoff compacts conversation state into a document for clean transitions between sessions. write-a-skill scaffolds a new skill with correct structure so you can extend the set without starting from scratch.
The miscellaneous layer adds git-guardrails-claude-code which blocks dangerous git operations, plus helpers for test migration, exercise scaffolding, and pre-commit hook setup.
Get the weekly deep dive
Tutorials on Claude Code, AI agents, and dev tools - delivered free every week.
From the archive
May 12, 2026 • 8 min read
May 12, 2026 • 9 min read
May 11, 2026 • 7 min read
May 10, 2026 • 8 min read
The install is a single command:
npx skills@latest add mattpocock/skills
The CLI presents an interactive selector so you can choose which skills to pull down rather than importing the full set at once. After selecting, open Claude Code (or your preferred agent tool) and run:
/setup-matt-pocock-skills
That slash command configures the skill for your repository. It asks which issue tracker you use, sets up triage labels, and records the location of your documentation. Everything after that follows the individual skill's own prompts.
The skills are written in Shell and work with any model - Claude Code, Codex, Gemini CLI, or any agent that supports a local skill directory. There are no API keys, no external services, and no cloud dependency. The skills run entirely from your local .claude directory or equivalent.
This repo is aimed at developers who are already using an AI coding agent daily and have hit the point where raw prompting stops scaling.
If you recognize any of these patterns, the skills are worth trying:
The skills encode solutions to each of those problems as repeatable, structured workflows. grill-with-docs exists because requirements interviews produce better context than raw prompts. diagnose exists because debugging needs a protocol, not intuition. handoff exists because context windows end and work should not.
Even developers who prefer Cursor or Windsurf over Claude Code will find the underlying patterns transferable. The engineering reasoning is not model-specific - it is about encoding judgment into repeatable steps that any agent can follow.
If you follow the Claude Code skills space through Developers Digest, this repo fits directly into the patterns catalogued at skills.developersdigest.tech. That directory covers community and first-party skills across categories like debugging, architecture, and agent productivity - the same territory mattpocock/skills occupies.
The four failure modes Pocock identifies - misalignment, verbosity, broken feedback loops, poor architecture - map closely to the design principles behind the structured agent workflows covered in DevDigest guides. The diagnose skill mirrors the debugging patterns from guides on building reliable agents. grill-with-docs is an applied version of the context-first approach covered in posts on CLAUDE.md patterns and session continuity.
If you are building your own agent workflow on top of Claude Code, the write-a-skill helper in this repo is worth a direct look. It produces a valid skill scaffold that you can immediately customize rather than starting from an empty file - the same way the DevDigest skill templates at hooks.developersdigest.tech provide starting points for hook-driven automation.
The handoff skill also parallels patterns used in production agent pipelines where conversation state must be preserved across long sessions. The underlying idea - compact what matters, discard what does not - is a pattern DevDigest has covered repeatedly as context windows become the primary constraint in agent-assisted development.
The star count reflects genuine community validation. But a few caveats are worth stating clearly before you install.
The skills deliver most value to developers already comfortable with their agent tool of choice. Someone still learning Claude Code basics will find the structured workflows add cognitive load before they deliver returns. grill-with-docs is most useful once you already know what CONTEXT.md and ADRs are for - and once you have a project large enough that architectural context matters.
The installation flow is smooth for Claude Code users. Other agent integrations - Cursor, Windsurf, Gemini CLI - require reading separate setup documentation, and the quality of that documentation varies by tool.
The caveman skill, which compresses output to cut token usage, works well in long internal sessions but produces output that can be harder to review or share with non-technical collaborators. It is a toggle, not a default, which is the right design choice.
The repo comes directly from Pocock's personal workflow, which is a genuine strength - these skills were selected by someone using them under real production conditions - but it also means the set reflects one developer's workflow preferences. Your most useful configuration will likely be a subset of what is here, customized for your own project conventions.
Overall, mattpocock/skills is a well-structured and immediately deployable set of engineering workflows. The decision to ship from a personal .claude directory gives it a level of real-world testing that most community skill collections lack.
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.
Anthropic's agentic coding CLI. Runs in your terminal, edits files autonomously, spawns sub-agents, and maintains memory...
View ToolGoogle's open-source coding CLI. Free tier with Gemini 2.5 Pro. Supports tool use, file editing, shell commands. 1M toke...
View ToolOpen-source AI pair programming in your terminal. Works with any LLM - Claude, GPT, Gemini, local models. Git-aware ed...
View ToolOpen-source AI code assistant for VS Code and JetBrains. Bring your own model - local or API. Tab autocomplete, chat,...
View ToolTurn a one-liner into a working Claude Code skill. From idea to installed in a minute.
View AppUnlock pro skills and share private collections with your team.
View AppCatch broken SKILL.md files in CI before they hit your team.
View AppConfigure model, tools, MCP, skills, memory, and scoping.
Claude CodeConnect external tools and data sources via the open MCP standard.
Claude CodePrefix prompts with ! to run shell commands directly, bypassing Claude.
Claude Code
Nimbalyst Demo: A Visual Workspace for Codex + Claude Code with Kanban, Plans, and AI Commits Try it: https://nimbalyst.com/ Star Repo Here: https://github.com/Nimbalyst/nimbalyst This video demos N...

Composio: Connect AI Agents to 1,000+ Apps via CLI (Gmail, Google Docs/Sheets, Hacker News Workflows) Check out Composio here: http://dashboard.composio.dev/?utm_source=Youtube&utm_channel=0426&utm_...

Anthropic has released Channels for Claude Code, enabling external events (CI alerts, production errors, PR comments, Discord/Telegram messages, webhooks, cron jobs, logs, and monitoring signals) to b...
obra/superpowers picked up 1,600+ GitHub stars in a single day by solving the real problem with coding agents - they ski...
Matt Pocock's Claude Code skills library picked up 3,372 GitHub stars on a single day. Here's what's inside, how to inst...
Matt Pocock released production-ready agent skills straight from his .claude directory. With 66k stars and 16k added thi...
The mattpocock/skills repository hit 62k stars on GitHub's weekly trending list with 17 composable Claude Code agent ski...
obra/superpowers is a composable, opinionated methodology for AI coding agents - structured workflows covering planning,...
A single CLAUDE.md file inspired by Andrej Karpathy's observations about LLM coding mistakes gained 24,000 GitHub stars...

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.