What This Repo Claims
23 opinionated tools that turn Claude Code into a virtual engineering team — a CEO who rethinks the product, an eng manager who locks architecture, a designer who catches AI slop, a reviewer who finds production bugs, a QA lead who opens a real browser, a security officer who runs OWASP + STRIDE audits.
Built by Garry Tan, President & CEO of Y Combinator. He claims to have shipped 600,000+ lines of production code in 60 days using this setup while running YC full time.
88.4k stars. 13k forks. One of the fastest-growing developer tool repos on GitHub in 2026.
Install command:
git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git
~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setup
What I Tested
Unlike reviews #1 and #2 on this site — which used controlled before/after tests on a clean environment — this review is based on real production usage. I used gstack throughout the entire build of repoverifier.dev, the site you're reading right now.
Environment:
- macOS, MacBook Pro 14-inch M4 Pro, 24GB RAM
- Claude Code 2.1.126
- gstack installed globally
What I built using gstack:
- repoverifier.dev — full-stack Node.js + Express + EJS + PostgreSQL
- 2 Railway services (website + cms)
- v2 features: GA4 analytics, like/dislike feedback, Resend email notifications
- Complete SEO implementation
Results — Skill by Skill
/plan-eng-review
Used to design the repoverifier.dev architecture before writing a single line of code.
Output: 632-line engineering review covering complete file structure for both services, all routes, full database schema with migration SQL, all environment variables, Railway config, inter-service communication pattern, and 10 explicit decisions requiring approval.
Every decision was documented and required a yes/no answer before the build started. Hidden assumptions were forced into the open.
The eng review caught the CMS pagination question, the slug generation approach, and the site_config editing scope — all before any code existed. This is the skill's core value: it finds the decisions you didn't know you needed to make.
/feature-dev
Used to build both services (website and cms) from the approved engineering review.
Result: 16 curl tests passed on first build. All routes working.
Both services deployed to Railway cleanly.
The build followed the eng review exactly. No scope drift. No undocumented additions.
/review
Run after /feature-dev. Caught a race condition in the vote submission logic before it ever touched production.
Fix applied: UNIQUE constraint + ON CONFLICT DO NOTHING on the feedback table. The review found a real bug that would have caused duplicate votes under concurrent load.
This is the moment gstack earns its keep. A solo developer would have shipped the race condition. /review caught it.
/qa
Run after v2 features (GA4, like/dislike feedback, Resend email).
Result: 20/20 tests passed. Caught issues before they reached Railway — including a formatting bug in the Markdown renderer and an edge case in the Resend client initialization.
/ship
Used for Railway deployment. Handled the git commit and push sequence cleanly across multiple deploys.
Install Experience
Single command install. Worked first time. No errors.
gstack is installed as skills under
~/.claude/skills/gstack/ — nothing touches PATH, nothing runs in the background. Clean.
Issues Documented
One non-blocking hook error appeared during usage:
Failed with non-blocking status code: [Hook] legacy hook execution failed for pre:edit-write:suggest-compact: spawnSync /usr/local/bin/node EPIPE
Appeared during /feature-dev on the repoverifier v2 build. Did not affect functionality. The word "legacy" in the hook name suggests this is a known transitional issue.
What I Did Not Test
- /plan-ceo-review — the product strategy skill
- /browse — the browser automation skill (requires Chrome setup)
- /cso — the security audit skill
- /retro — the retrospective skill
- GBrain integration
gstack has 23+ skills. This review covers the core development workflow: plan → build → review → qa → ship. The remaining skills require longer-term use to evaluate properly.
Verdict: Solid
The install works. The workflow is coherent end to end. The skills do what they claim — /plan-eng-review surfaces hidden decisions, /review catches real bugs, /qa finds edge cases before production.
The Garry Tan credibility claim is unverifiable from the outside. Whether he shipped 600k lines in 60 days is irrelevant to whether the workflow improves your output. Based on a real production build, it does.
The one caveat: gstack's value compounds over time and across projects. A single session review undersells it. A developer who uses it consistently will get more value than this review captures.
Worth installing. Run /plan-eng-review on your next feature before writing a line of code.