gstack vs mattpocock/skills: How Garry Tan's Claude Code Setup Stacks Up Against the Community Standard (2026)

Garry Tan's gstack hit 66,000 GitHub Stars in weeks — MIT licensed, free. It turns Claude Code into a "virtual engineering team" with 23 slash commands covering the full product lifecycle from planning to deployment to retrospectives. This hands-on comparison with mattpocock/skills helps you decide which one to install first.

What Is gstack? The Core Value in 3 Minutes

gstack (GitHub: garrytan/gstack) was open-sourced by Y Combinator CEO Garry Tan in March 2026. 66,000+ Stars, MIT license, 35 contributors — with a significant portion of commits co-authored with Claude Opus 4.6.

Unlike mattpocock/skills (a general-purpose toolkit for devops, debugging, and testing), gstack is opinionated. Every command ships with a fixed R&D workflow baked in.

Install in one line:

git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setup

The setup script auto-generates a CLAUDE.md snippet in ~/.claude/. This file tells Claude Code to use gstack's /browse for all web browsing tasks, never call mcp__chrome__* tools directly, and exposes the available slash commands: /office-hours, /plan-ceo-review, /plan-eng-review, /plan-design-review, /design-consultation, /design-shotgun, /design-html, /review, /ship, /land-and-deploy, /canary, /benchmark, /browse, /connect-chrome, /qa, /qa-only, /design-review, /setup-browser-cookies, /setup-deploy, /setup-gbrain, /retro, /investigate, /document-release, /document-generate, /codex, /cso, /autoplan, /plan-devex-review, /devex-review, /careful, /freeze, /guard, /unfreeze, /gstack-upgrade, /learn.

5 Real gstack Workflows (Tested)

/review: Code Review in 3 Minutes

/review

Point it at a project path and Claude Code analyzes code quality, security risks, and style issues, outputting a structured review report. In testing on a TypeScript monorepo, one review covered 8 files in ~90 seconds. mattpocock/skills has devops/review — but it's a thin wrapper around existing linters, not a full Claude-native review workflow. gstack's /review is purpose-built for how Claude Code actually works.

/ship: Deploy to Fly.io in One Command

/ship

This triggers a full deployment pipeline: environment check → build → push to remote → startup verification. In testing, /ship defaults to Fly.io as the target. If you use Railway or Vercel you'll need to specify --target railway manually — a detail gstack's docs acknowledge but the CLAUDE.md doesn't set as a default.

/qa: Quality Assurance, Coverage and Regression

/qa

Runs unit tests, integration tests, checks coverage, and compares against the last known good build to catch regressions. mattpocock/skills' testing/test command provides basic test execution — no coverage analysis, no regression comparison. gstack's /qa feels closer to what an actual QA engineer would hand you.

/benchmark: Performance Baseline Testing

/benchmark

Runs local performance benchmarks, records key metrics (latency, throughput, memory), and generates a comparison report against previous runs. mattpocock/skills has no equivalent. For backend services where performance is a recurring concern, /benchmark can replace hand-rolled wrk or hey scripts.

/investigate: Deep Debugging

/investigate

Point it at an error log or anomalous behavior and Claude Code enters root-cause analysis mode: traces call chains, logs variables, and proposes fixes. In testing, this command worked best on Node.js projects (because mcp__node__* tools are naturally compatible with gstack's debugging flow). For Python projects, you need to manually specify the venv path.

gstack vs mattpocock/skills: Head-to-Head

**Privacy flag**: A Hacker News commenter noted that gstack contains telemetry on usage data. I could not locate an official gstack telemetry disclosure page — I recommend auditing ~/.claude/skills/gstack/ for send-metrics.* or telemetry.* files before installing on sensitive projects. If you handle proprietary or regulated code, this is the first thing to check.

Which One to Install? Decision Tree

Need a full product workflow (plan → ship → retrospective)?
├── Yes → gstack (ship+qa+investigate chain)
└── No → mattpocock/skills (wider tool coverage)

Privacy or compliance requirements (enterprise intranet / medical / financial code)?
├── Yes → mattpocock/skills (gstack telemetry concern)
└── No → Both are worth trying

Need a built-in benchmark command?
├── Yes → gstack (no mattpocock/skills equivalent)
└── No → mattpocock/skills (broader general coverage)

TL;DR

gstack and mattpocock/skills are not the same tool — gstack is for developers who want Claude Code to run the entire product, mattpocock/skills is for developers who want Claude Code to solve specific problems better. If you're a solo developer or small team, gstack's /ship+/qa+/investigate combo saves at least 30 minutes/day on deployment and debugging, provided you're comfortable with its telemetry design. If you have strict privacy requirements, mattpocock/skills is the safer starting point.

Related Reading

👉 Join MiniMax Token Plan: AI coding acceleration for businesses

👉 Join Zhipu Coding Plan: GLM-4.6/GLM-5 coding packages, China-stable, pay-per-token unlimited

👉 Join Aliyun AI: Top AI products with exclusive coupons for business innovation

📌 This article was AI-assisted generated and human-reviewed | TechPassive — An AI-driven content testing site focused on real tool reviews