Wanda

Posted on Mar 23 • Originally published at apidog.com

What Is gstack? This will Change How you Code Forever

TL;DR / Quick Answer

gstack is Garry Tan’s open-source system that transforms Claude Code into a virtual engineering team of 20 specialists. With structured slash commands, Garry ships 10,000-20,000 lines of production code per day (35% tests) while running YC full-time. Use commands like /office-hours for product strategy, /plan-ceo-review for scope validation, /review for bug detection, /qa for browser testing, and /ship for deployment. Install it in 30 seconds:

git clone ~/.claude/skills/gstack && ./setup

It’s free and MIT licensed.

Try Apidog today

Introduction

When Andrej Karpathy said on the No Priors podcast in March 2026, “I don’t think I’ve typed like a line of code probably since December, basically, which is an extremely large change,” the tech world listened. The paradigm has shifted: one developer with AI agents can ship like a team.

Peter Steinberger built OpenClaw (247,000 GitHub stars) mostly solo using AI agents. Garry Tan, President and CEO of Y Combinator (YC), works with startups like Coinbase, Instacart, and Rippling, and brings decades of engineering experience. gstack is his answer: How can one person ship like a team of twenty?

In 60 days, Garry shipped over 600,000 lines of production code (35% tests), averaging 10,000-20,000 lines per day, part-time. His last /retro across 3 projects showed 140,751 lines added, 362 commits, and ~115k net LOC in a single week. The difference? Tooling.

This guide gives you the actionable steps to install, use, and optimize gstack in your workflow. Includes skill catalog, real examples, and usage philosophy.

💡 If you’re building API products, gstack integrates naturally with Apidog for API testing and documentation workflows. The /qa skill can validate API endpoints, and /document-release keeps your API docs synced with shipped changes.

What Is gstack?

gstack is an open-source suite of 28 Claude Code skills that turn Claude into a virtual engineering team—each skill is a specialist (CEO, eng manager, designer, reviewer, QA lead, security officer, release engineer, etc.). All invoked as slash commands in Markdown. MIT license, no cost.

The Core Insight

Most AI coding tools treat you as a solo developer. gstack treats you as a CEO with a team. When you ask for a feature, gstack starts with /office-hours—a structured consultation, then moves through scope review, architecture, design, code review, QA, and deployment.

The Sprint Structure

gstack follows a strict process:

Think → Plan → Build → Review → Test → Ship → Reflect

Each skill hands off to the next. Design docs flow into reviews, test plans feed into QA, bugs discovered are verified as fixed, ensuring full coverage.

The 28 Skills Explained

Product & Strategy Skills

`/office-hours` — YC Office Hours

Specialist: YC Partner
Purpose: Start every project with six structured questions. Reframe, challenge, and generate implementation alternatives.

Example output:

You said "daily briefing app." But what you actually described is a
personal chief of staff AI. Here are 5 capabilities you didn't realize
you were describing...

[challenges 4 premises — you agree, disagree, or adjust]
[generates 3 implementation approaches with effort estimates]

RECOMMENDATION: Ship the narrowest wedge tomorrow, learn from real usage.

When: First skill on any new feature or product.

`/plan-ceo-review` — CEO / Founder

Specialist: CEO
Purpose: Rethink from first principles. Find the 10-star product, challenge or reduce scope.
When: After /office-hours, before implementation.

`/plan-design-review` — Senior Designer

Specialist: Product Designer
Purpose: Score each design dimension, suggest improvements, includes AI slop detection.
When: After eng review, before implementation.

`/design-consultation` — Design Partner

Specialist: Design Partner
Purpose: Build a complete design system, generate mockups.
When: When a full design system is needed.

Engineering & Architecture Skills

`/plan-eng-review` — Eng Manager

Specialist: Eng Manager
Purpose: Lock architecture, data flow, edge cases, tests. Generates ASCII diagrams.

Example:

┌─────────────┐     ┌──────────────┐     ┌────────────┐
│   Client    │────▶│  API Gateway │────▶│  Database  │
└─────────────┘     └──────────────┘     └────────────┘
       │                    │
       ▼                    ▼
  [State Cache]      [Rate Limiter]

When: After CEO/design review, pre-coding.

`/review` — Staff Engineer

Specialist: Staff Engineer
Purpose: Find/fix bugs that pass CI but fail in production. Flags completeness issues.

Example:

[AUTO-FIXED] 2 issues:
- Null check missing in getUserById()
- Unhandled promise rejection in api handler

[ASK] Race condition in concurrent update → you approve fix
[COMPLETENESS GAP] No retry logic for transient failures

When: After implementation, before /qa.

`/investigate` — Debugger

Specialist: Debugger
Purpose: Root-cause debug, stops after 3 failed fixes.
When: When /review can't auto-fix a bug.

`/codex` — Second Opinion

Specialist: OpenAI Codex
Purpose: Independent code review, adversarial challenge, or open consultation.
When: After /review, for cross-model validation.

Testing & QA Skills

`/qa` — QA Lead

Specialist: QA Engineer
Purpose: Open real browser, click through flows, auto-fix bugs, generate regression tests.

Workflow:

1. Opens staging URL in Chromium
2. Runs test plan from /plan-eng-review
3. Finds bug and fixes with atomic commit
4. Re-verifies and generates regression test

When: After /review, on staging URL.

`/qa-only` — QA Reporter

Specialist: QA Reporter
Purpose: Bug report only, no code changes.
When: For audit trails, or when fixes are handled elsewhere.

`/benchmark` — Performance Engineer

Specialist: Performance Engineer
Purpose: Track FCP, LCP, CLS, TTI, bundle sizes on PRs.
When: Before/after refactors, performance PRs.

`/browse` — QA Engineer

Specialist: Browser Automation
Purpose: Real Chromium browser, real interactions.
Commands: goto <url>, click <selector>, type <selector> <text>, screenshot <name>, wait <selector>

`/setup-browser-cookies` — Session Manager

Specialist: Browser Session Manager
Purpose: Import cookies for authenticated page testing.
When: Before /qa on login-required apps.

Security & Compliance Skills

`/cso` — Chief Security Officer

Specialist: Security Officer
Purpose: Run OWASP Top 10 + STRIDE audit, exclude false positives, show exploit scenarios.

Example:

[CRITICAL] SQL Injection in /api/users?id=
Exploit: GET /api/users?id=1' OR '1'='1
Impact: Full DB access
Fix: Parameterized queries

When: Before production releases, user data/auth handling features.

Shipping & Deployment Skills

`/ship` — Release Engineer

Specialist: Release Engineer
Purpose: Sync main, run tests, audit coverage, open PR.
Workflow:
1. Checkout main, pull
2. Create feature branch
3. Run tests (bootstrap if missing)
4. Audit coverage
5. Push and open PR

`/land-and-deploy` — Deployment Engineer

Specialist: Deployment Engineer
Purpose: Merge PR, wait for CI, deploy, verify prod health.
When: After PR approval.

`/canary` — SRE

Specialist: Site Reliability Engineer
Purpose: Monitor post-deploy for errors, regressions, failures.

`/document-release` — Technical Writer

Specialist: Technical Writer
Purpose: Update docs to match shipped changes.
When: After /ship or /land-and-deploy.

Reflection & Analytics Skills

`/retro` — Eng Manager

Specialist: Eng Manager
Purpose: Weekly retro, per-person/project breakdowns, test trends, shipping streaks.

Power Tools

/careful: Warn before destructive commands.
/freeze: Lock edits to a directory.
/guard: Full safety mode (/careful + /freeze).
/unfreeze: Remove edit lock.
/setup-deploy: One-time deploy config.
/autoplan: Run CEO → design → eng review automatically.
/gstack-upgrade: Self-updater.

Installation Guide

gstack installs in 30 seconds. It does not modify your PATH or run in the background. Everything is self-contained in .claude/.

Requirements

Claude Code
Git
Bun v1.0+
Node.js (Windows only)

Step 1: Install on Your Machine

Open Claude Code and run:

git clone https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setup

Add a “gstack” section to CLAUDE.md:

Use /browse from gstack for all web browsing.
Never use mcp__claude-in-chrome__* tools.
List available skills.

Step 2: Add to Your Repo (Optional)

To share with teammates:

cp -Rf ~/.claude/skills/gstack .claude/skills/gstack && rm -rf .claude/skills/gstack/.git && cd .claude/skills/gstack && ./setup

Update the project’s CLAUDE.md as above.

Codex, Gemini CLI, or Cursor

gstack works with any agent supporting the SKILL.md standard. Skills go in .agents/skills/.

Install to one repo:

git clone https://github.com/garrytan/gstack.git .agents/skills/gstack
cd .agents/skills/gstack && ./setup --host codex

User-wide install:

git clone https://github.com/garrytan/gstack.git ~/gstack
cd ~/gstack && ./setup --host codex

Auto-detect agents:

git clone https://github.com/garrytan/gstack.git ~/gstack
cd ~/gstack && ./setup --host auto

Troubleshooting

Skill not showing up?
```
cd ~/.claude/skills/gstack && ./setup
```

/browse fails?

cd ~/.claude/skills/gstack && bun install && bun run build

Stale install?

Run /gstack-upgrade or set auto_upgrade: true in ~/.gstack/config.yaml.
Windows users:

Use Git Bash or WSL. Ensure bun and node are on PATH.

Claude can’t see skills?

Add this to your project’s CLAUDE.md:

## gstack
Use /browse from gstack for all web browsing. Never use mcp__claude-in-chrome__* tools.
Available skills: /office-hours, /plan-ceo-review, /plan-eng-review, /plan-design-review,
/design-consultation, /review, /ship, /land-and-deploy, /canary, /benchmark, /browse,
/qa, /qa-only, /design-review, /setup-browser-cookies, /setup-deploy, /retro,
/investigate, /document-release, /codex, /cso, /autoplan, /careful, /freeze, /guard,
/unfreeze, /gstack-upgrade.

Real-World Example: Building a Daily Briefing App

Here’s a practical gstack workflow:

You:    I want to build a daily briefing app for my calendar.
You:    /office-hours
Claude: [asks about the pain — specific examples, not hypotheticals]

You:    Multiple Google calendars, events with stale info, wrong locations...

Claude: [reframes: you're building a personal chief of staff AI]
        [challenges premises, suggests approaches, writes design doc]

You:    /plan-ceo-review
You:    /plan-eng-review
        [ASCII diagrams, test matrix, security concerns]
You:    Approve plan.
        [writes code]
You:    /review
        [auto-fixes issues, flags gaps]
You:    /qa https://staging.myapp.com
        [browser testing, fixes a bug]
You:    /ship
        [runs tests, opens PR]

Eight commands—product idea to PR with CI, QA, and docs.

Parallel Sprints: Running Multiple Agents at Once

gstack can run multiple sprints in parallel. Conductor launches several Claude Code sessions in isolated workspaces:

Session 1: /office-hours (refining product spec)
Session 2: /review (reviewing features)
Session 3: /qa (testing deployment)
Session 4: Implementation

With process, parallel agents work without chaos.

Privacy & Telemetry

gstack features opt-in telemetry:

Default: off. No data sent unless you opt in.
On first run: prompts to share anonymous usage data.
If enabled: sends skill name, duration, success/fail, version, OS. Never any code, paths, repo names, or user content.
Turn off: gstack-config set telemetry off
Data stored in Supabase (open source Firebase alternative).
Local analytics: run gstack-analytics for private dashboard from local JSONL file.

Who Should Use gstack?

Founders/CEOs: Ship at startup speed solo.
First-time Claude Code users: Structured roles, guided prompts.
Tech leads/staff engineers: Rigorous review & QA, automation on every PR.
Solo builders: Virtual team workflow.
YC startups: Built for YC founders.

Who Should Skip gstack?

Established teams: Already have CI/CD, review, and design systems? Use individual skills as needed.
Non-Claude Code users: Best with Claude; works with Codex, Gemini CLI, Cursor (via SKILL.md).
Freeform AI explorers: Prefer unstructured prompt engineering? gstack’s structure may feel limiting.

The Philosophy Behind gstack

Boil the lake: Commit fully or not at all.
Search before building: Reuse before reinventing.
Three layers of knowledge: Explicit (docs), tacit (intuition), unknown (blind spots). gstack encodes tacit into explicit via skills.
Iron Law of Debugging: No fixes without root-cause investigation. Three failed fixes? Stop and reassess.

Conclusion

gstack is Garry Tan’s answer to “How does one person ship like a team of twenty?” The secret: not working harder, but working with better tooling. Use structured slash commands to bring specialists into your Claude Code workflow.

Typical sprint:

/office-hours — reframe the problem
/plan-ceo-review — challenge scope
/plan-eng-review — lock architecture
/plan-design-review — rate design
Build — implement
/review — catch bugs
/qa — browser test
/ship — push PR

Next steps:

Install gstack:

   git clone ~/.claude/skills/gstack && ./setup

Run /office-hours on your next idea.
See if the output changes your approach.

FAQ

Is gstack free?

Yes, 100% free and MIT licensed. No premium tier.

Do I need Claude Code to use gstack?

Optimized for Claude Code, works on any SKILL.md agent: Codex CLI, Gemini CLI, Cursor.

How long does installation take?

About 30 seconds. Clone, run ./setup.

Can I use individual skills?

Yes. Use /review and /qa independently if desired.

Does gstack work with private repos?

Yes. Skills live in .claude/skills/gstack. Commit to share with teammates.

Windows support?

Works on Windows 11 via Git Bash/WSL. Requires Node.js and Bun.

How does `/browse` work?

Uses Playwright to control headless Chromium. Use /setup-browser-cookies for authenticated sessions.

Can I customize skills?

Yes. Skills are Markdown files; edit as needed, PR upstream improvements.

Difference between `/qa` and `/qa-only`?

/qa finds and fixes, /qa-only only reports.

How does telemetry work?

Opt-in only, no code or user data sent. Disable anytime with gstack-config set telemetry off.

What if I hit a gstack bug?

Run /investigate on gstack’s codebase or open a GitHub issue.

Can I run gstack skills in parallel?

Yes, with Conductor. Multiple sessions, isolated workspaces.

TL;DR / Quick Answer

Introduction

What Is gstack?

The Core Insight

The Sprint Structure

The 28 Skills Explained

Product & Strategy Skills

/office-hours — YC Office Hours

/plan-ceo-review — CEO / Founder

/plan-design-review — Senior Designer

/design-consultation — Design Partner

Engineering & Architecture Skills

/plan-eng-review — Eng Manager

/review — Staff Engineer

/investigate — Debugger

/codex — Second Opinion

Testing & QA Skills

/qa — QA Lead

/qa-only — QA Reporter

/benchmark — Performance Engineer

/browse — QA Engineer

/setup-browser-cookies — Session Manager

Security & Compliance Skills

/cso — Chief Security Officer

Shipping & Deployment Skills

/ship — Release Engineer

/land-and-deploy — Deployment Engineer

/canary — SRE

/document-release — Technical Writer

Reflection & Analytics Skills

/retro — Eng Manager

Power Tools

Installation Guide

Requirements

Step 1: Install on Your Machine

Step 2: Add to Your Repo (Optional)

Codex, Gemini CLI, or Cursor

Troubleshooting

Real-World Example: Building a Daily Briefing App

Parallel Sprints: Running Multiple Agents at Once

Privacy & Telemetry

Who Should Use gstack?

Who Should Skip gstack?

The Philosophy Behind gstack

Conclusion

FAQ

Is gstack free?

Do I need Claude Code to use gstack?

How long does installation take?

Can I use individual skills?

Does gstack work with private repos?

Windows support?

How does /browse work?

Can I customize skills?

Difference between /qa and /qa-only?

How does telemetry work?

What if I hit a gstack bug?

Can I run gstack skills in parallel?

`/office-hours` — YC Office Hours

`/plan-ceo-review` — CEO / Founder

`/plan-design-review` — Senior Designer

`/design-consultation` — Design Partner

`/plan-eng-review` — Eng Manager

`/review` — Staff Engineer

`/investigate` — Debugger

`/codex` — Second Opinion

`/qa` — QA Lead

`/qa-only` — QA Reporter

`/benchmark` — Performance Engineer

`/browse` — QA Engineer

`/setup-browser-cookies` — Session Manager

`/cso` — Chief Security Officer

`/ship` — Release Engineer

`/land-and-deploy` — Deployment Engineer

`/canary` — SRE

`/document-release` — Technical Writer

`/retro` — Eng Manager

How does `/browse` work?

Difference between `/qa` and `/qa-only`?