Matt Pocock's skills repo had climbed to 45,289 GitHub stars in a late-Wednesday API snapshot after his Claude Code workflow files spread through YouTube and open-source discovery posts. Created February 3 and pushed again April 29, the MIT-licensed project packages reusable agent instructions as markdown SKILL.md files and installer targets. The project places private prompt routines in a public, versioned repository.

Key Takeaways

AI-generated summary, reviewed by an editor. More on our AI guidelines.

GitHub numbers moved in hours

An Indie Hacker News video published Wednesday put the repo at 36,800 stars and about 2,800 forks. A later GitHub API snapshot showed 45,289 stars, 3,652 forks, 371 subscribers and 13 open issues, turning the video's count into a same-day floor rather than the current figure.

The repo's README states the premise directly: Pocock uses the skills for "real engineering" rather than vibe coding. It argues that heavier systems such as GSD, BMAD and Spec-Kit can take control away from the engineer, while small composable prompts keep the process visible.

Comparing those snapshots, the repo added 8,489 stars, equal to roughly 23 percent of the video's total.

The package is mostly process

A public tree scan found 22 SKILL.md entries, including personal and deprecated folders, and 18 after excluding deprecated entries. The Claude plugin manifest exposes 12 skills, including grill-me, tdd, diagnose, to-prd, to-issues, triage, zoom-out and caveman.

Pocock's quickstart uses npx skills@latest add mattpocock/skills and does not ship a separate agent runtime. The setup step asks users to run /setup-matt-pocock-skills so the local project records its issue tracker, triage labels and documentation layout.

Raw examples show the design, with grill-me telling the agent to ask one question at a time until a plan has been tested against each branch of the decision tree. tdd rejects bulk test-writing and directs the agent through one failing test, one minimal fix and a repeat loop.

Pocock had previewed the method

In a March 16 AI Hero article, Pocock wrote that AI agents act like capable engineers with one problem: they have no memory. His answer was not a larger context dump, but repeatable process files for planning, requirements, test cycles and codebase architecture checks.

That earlier article described a short grill-me session producing 16 questions and longer sessions producing 30 to 50. It also framed to-issues as a way to break a plan into vertical slices, so a ticket covers a narrow path through schema, API, interface and tests instead of one horizontal layer.

The trade-off is maintenance

The repo's sudden popularity does not make Pocock's preferences universal. The GitHub tree includes personal and deprecated folders, and the manifest narrows the active plugin list. The npm package used by the installer is maintained under vercel-labs/skills, while Pocock's repo is a source collection consumed by that installer path.

That installation split matters for teams copying the pattern because a skill instruction can make an agent ask better questions, while also becoming another artifact that needs review when project conventions change. In Pocock's release, the reusable object is not only the prompt, but the document that tells the prompt when to run.

Frequently Asked Questions

What is Matt Pocock's skills repo?

It is a public GitHub repository containing markdown-based agent instructions from Pocock's Claude Code workflow. The files define repeatable behaviors such as grilling a plan, writing PRDs, breaking work into issues and running TDD loops.

How many GitHub stars did the repo have?

A GitHub API snapshot late Wednesday showed 45,289 stars and 3,652 forks. Earlier videos and writeups used lower counts because the repository was still gaining stars during the same news cycle.

Does the repo ship an AI coding agent?

No. The repo is mostly process instructions, not a runtime. Its README points users to the npm skills installer and then to a setup skill that records project-specific issue tracker and documentation choices.

How many skills are active in the plugin manifest?

The repository contains 22 SKILL.md files when personal and deprecated folders are included. The Claude plugin manifest exposes 12 skill directories as the active install set.

Why does this matter for AI coding teams?

The repo shows prompts becoming maintained project artifacts. Teams can use skill files to encode review, testing and planning habits, but those files also need updates when project conventions change.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Repo Radar: 5 GitHub Projects Worth Your Week
GitHub's breakout projects this week are not just libraries. They are early products: agents that remember, recorders that challenge paid apps, local AI demos, tutor workspaces, and dashboards for man
Intermediate Tutorial: How To Build Reliable Workflows With OpenAI Codex
Most developers hit the same wall with Codex. The first week feels electric. You prompt, it codes, things work. Then the codebase grows. Fixes start landing in the wrong layer. Architecture quietly de
Beginner Tutorial: How to Use Claude Cowork. Seven Capabilities That Replace Your Busywork
Claude Cowork turned Anthropic's chatbot into a desktop agent back in January 2026. Three months in, the tool handles file management, persistent memory, third-party connectors, reusable skills, proje
Tools & Workflows

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]