R&D Management 11 min read

What Makes the Matt Pocock ‘Skills’ Repo Reach 230K Stars and 18M Installs?

The article examines Matt Pocock’s open‑source “skills” repository—its structure, key commands, real‑world workflow integrations, installation steps, and practical scenarios—showing why it has amassed over 230 000 stars and 18 million installations among AI‑assisted developers.

Java Companion
Java Companion
Java Companion
What Makes the Matt Pocock ‘Skills’ Repo Reach 230K Stars and 18M Installs?

Background and Motivation

The author shares personal frustration with AI‑generated code that often requires rework because requirements were not fully clarified. To address this, Matt Pocock’s publicly released mattpocock/skills repository is introduced as a set of reusable “skill” files that encode disciplined development practices for AI agents.

What the Repository Contains

Pocock, known for the Total TypeScript project and former engineer at Vercel and Stately, open‑sourced the .agents directory from his daily workflow. A “skill” is a folder with a markdown description that tells the AI how to perform a specific task; it is invoked with a command such as /tdd. Unlike static project docs, a skill loads only for the conversation that triggers it, keeping the context lightweight.

The repo now has 230 000+ stars, 2 000+ forks, and more than 18 million installations across 53 listed skills.

Core Skills and Their Functions

/grill-me

and /grill-with-docs are the most popular, together installed over 900 000 times. They reverse the usual AI‑to‑human flow: the AI interrogates the user to surface hidden requirement branches, asking one question at a time until every path is answered. /grill-with-docs additionally aggregates terminology into a CONTEXT.md file and records irreversible technical decisions as ADR documents. /ask-matt lets users describe a situation and receive a skill recommendation from the AI. /tdd enforces test‑driven development by requiring a failing test before any implementation, preventing the “fake TDD” where all tests are written up front. /diagnosing-bugs mandates a reproducible command before any hypothesis is formed, and /code-review separates style checks from functional verification. /improve-codebase-architecture scans the repository for “deep modules, shallow modules” patterns, generating an HTML report of candidates for refactoring; the user decides which to address.

Composing End‑to‑End Workflows

/to-spec

: converts a finished conversation into a specification issue. /to-tickets: breaks the spec into ordered tickets with explicit dependencies. /implement: follows the spec or tickets, automatically invoking /tdd and /code-review during development. /triage: processes a backlog of issues through a state‑machine. /handoff: when a session becomes long or sluggish, it creates a handoff document and starts a fresh conversation. /prototype: builds a throw‑away prototype for ambiguous problems and discards it afterward.

Installation and Setup

For Claude Code users, a single command installs the plugin: claude plugins install mattpocock-skills Alternatively, the repository can be added with: npx skills@latest add mattpocock/skills After installation, run /setup-matt-pocock-skills once per repository to configure issue locations, triage labels, and documentation directories.

The skills are model‑agnostic; the README notes they work with any model that supports the Agent Skills specification.

When to Use and When Not to Use

New, vague requirements – the grill skills surface hidden assumptions.

Bug‑reproduction failures – /diagnosing-bugs forces a reproducible command.

Legacy codebases becoming tangled – periodic runs of /improve-codebase-architecture highlight refactor targets.

Non‑coding tasks such as decision‑making or product brainstorming – /grill-me can guide discussions.

Conversely, for quick demos or one‑off scripts, the overhead of the skill set outweighs its benefits; a simple AI code generation call is sufficient.

Author’s Assessment

The repository does not introduce novel concepts; it codifies decades‑old software‑engineering practices—requirement interrogation, test‑first development, unified terminology, ADR documentation, and prototype validation—into AI‑executable files. Its transparency (plain‑text skills) contrasts with black‑box frameworks that make all decisions for the user.

The author recommends installing only the most relevant skills, starting with /grill-with-docs, and gradually adding others as needed.

Open‑Source Link

https://github.com/mattpocock/skills
Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

code generationAI agentssoftware engineeringopen-sourceworkflow automationskills repository
Java Companion
Written by

Java Companion

A highly professional Java public account

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.