Build This Now
Build This Now
Qu'est-ce que le code Claude ?Installer Claude CodeL'installateur natif de Claude CodeTon premier projet Claude Code
Principes de base de l'agentAgents en arrière-plan dans Claude CodeRoutage des sous-agentsConception de sous-agents dans Claude CodeDistribution de tâches dans Claude CodeÉquipes d'agents Builder-ValidatorLes équipes d'agents Claude CodeContrôles des équipes d'agentsTemplates de prompts pour les équipes d'agentsMeilleures pratiques des équipes d'agentsWorkflow des équipes d'agentsAgents personnalisésPatterns d'agentsHow Garry Tan (YC CEO) Uses Claude Code: Inside the 23-Tool gstack SetupDes agents qui ressemblent à des humainsHermes Agent : l'IA qui s'améliore elle-mêmeL'ingénierie du harness agent
speedy_devvkoen_salo
Blog/Handbook/Agents/How Garry Tan (YC CEO) Uses Claude Code: Inside the 23-Tool gstack Setup

How Garry Tan (YC CEO) Uses Claude Code: Inside the 23-Tool gstack Setup

The 23 slash commands, six roles plus a CSO, and one install line behind Garry Tan's Claude Code setup. Plus the criticism.

Arrêtez de configurer. Commencez à construire.

Templates SaaS avec orchestration IA.

Published May 15, 202611 min readHandbook hubAgents index

Problem: You watched Garry Tan's launch tweet, skimmed the TechCrunch piece, maybe caught the SXSW clip, and now you want to know what is actually in gstack. The repo has 97k stars. The README is dense. Half of X is calling it god mode. The other half is calling it a folder of prompts. Neither side is showing you the table.

Quick Win: Clone gstack into your Claude Code skills folder with one line and you get 23 slash commands grouped under a Think to Ship loop:

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

This post lists every command, names every role, walks the workflow, handles the criticism honestly, and tells you where gstack stops and a production skeleton starts.

What gstack Is

gstack is an open-source pack of opinionated Claude Code skills published by Garry Tan, the President and CEO of Y Combinator. He shipped it on March 12, 2026 under MIT. Every slash command embodies a specialist persona with its own priorities, constraints, and outputs. The work moves through a fixed loop: Think, Plan, Build, Review, Test, Ship, Reflect.

Repo: github.com/garrytan/gstack. As of 2026-05-15, the README ships 23 core skills (the count grew from 6 at launch, to 13 by the TechCrunch coverage on March 17, to 23 now).

Stats today: 97.4k stars, 14.5k forks, 597 watchers, plus 915 stars in the last 24 hours alone. (source)

Why It Matters

YC has funded 4,000+ companies. The CEO running that engine is publishing the exact skill setup he uses to ship code. That is the rarest possible signal for a Claude Code workflow. You can argue with the design. You cannot argue with the source.

Garry stated his own throughput on the launch post:

"I just open-sourced my entire Claude Code setup I used to average 10K LOC and 100 PRs per week in the last 50 days." (LinkedIn, 2026-03-12)

Self-reported numbers. Frame them that way. The shape of the workflow is the part you can copy.

The Role-Based Architecture

gstack groups every skill under a job title. A role owns priorities, constraints, and outputs. Free-form prompting asks Claude to wear many hats inside one message. Roles split the hats and pass work between them.

RoleWhat it owns
CEO / FounderScope, product framing, what to build
DesignerVisual system, mockups, AI slop detection
Eng ManagerArchitecture, data flow, edge cases
Release ManagerSync, test, push, PR, deploy
Doc EngineerDocs in sync with shipped code
QA LeadReal browser testing, bug fixes
Chief Security OfficerOWASP Top 10 + STRIDE threat modeling

The loop reads as one sentence: Think (Office Hours plus CEO review), Plan (Eng plus Design plus DevEx review), Build, Review, Test (QA plus CSO), Ship, Reflect (retro). Source: gstack README.

The 23 Tools

The core inventory, sourced verbatim from the gstack README and docs/skills.md, accessed 2026-05-15.

#Slash CommandRoleWhat it does
1/office-hoursYC Office HoursSix forcing questions that reframe the product before any code
2/plan-ceo-reviewCEO / FounderScope review across four modes: expansion, selective, hold, reduction
3/plan-eng-reviewEng ManagerArchitecture, data flow, diagrams, edge cases, test matrix
4/plan-design-reviewSenior Designer0 to 10 ratings per design dimension; flags AI slop
5/plan-devex-reviewDX LeadInteractive developer-experience audit, three modes
6/design-consultationDesign PartnerEnd-to-end design system: research plus mockups
7/design-shotgunDesign Explorer4 to 6 AI mockup variants with taste-memory learning
8/design-htmlDesign EngineerMockup converted to production HTML, ~30KB, zero deps
9/reviewStaff EngineerProduction bug detection with auto-fixes and coverage audit
10/investigateDebuggerSystematic root-cause analysis, traces data flow, three-fix limit
11/design-reviewDesigner Who CodesPost-ship design audit and atomic-commit auto-fixes
12/devex-reviewDX TesterLive onboarding audit with timing and error screenshots
13/qaQA LeadBrowser testing, fixes bugs, generates regression tests
14/qa-onlyQA ReporterBug reports only, no code edits
15/csoChief Security OfficerOWASP Top 10 + STRIDE; 17 false-positive exclusions
16/shipRelease EngineerSync, test, audit, push, open PR; bootstraps frameworks
17/land-and-deployRelease EngineerMerge, run CI, deploy, verify production health
18/canarySREPost-deploy monitoring: console errors, perf, failures
19/benchmarkPerformance EngineerCore Web Vitals, resource sizes, before-and-after diffs
20/document-releaseTechnical WriterAuto-updates docs, Diataxis coverage map
21/document-generateDoc AuthorGenerates missing reference, how-to, and tutorial docs
22/retroEng ManagerWeekly retro, per-person breakdowns, streak tracking
23/browseQA EngineerReal Chromium browser, ~100ms per command

That is the 23-tool count locked to 2026-05-15. The README is moving fast. Pin the commit you cloned if you want a stable reference.

A Real Workflow End to End

A clean session moves through six commands in this order: think, plan, build, review, test, ship.

  1. /office-hours: Six forcing questions reframe the feature. You answer in plain English. Output is a sharper scope.
  2. /plan-ceo-review: A CEO persona checks scope across four modes (expand, selective, hold, reduce). Often kills work before it starts.
  3. /plan-eng-review: An Eng Manager persona drafts architecture, data flow, edge cases, and a test matrix.
  4. Implement the work as Claude normally would, against the plan.
  5. /review: A Staff Engineer persona scans for production bugs and applies atomic auto-fixes. Coverage audit runs in the same pass.
  6. /qa: A QA Lead persona opens a real Chromium browser, runs the flow, fixes bugs it finds, and writes regression tests.
  7. /ship: A Release Engineer persona syncs, tests, audits, pushes, and opens a PR in one chain.

Each command keeps Claude in one role for the duration of that step. That is the design choice the rest of the system is built on.

The Power Tools Most Posts Skip

Beyond the 23 there is a second tier of utility commands: /codex, /careful, /freeze, /guard, /unfreeze, /open-gstack-browser, /setup-deploy, /gstack-upgrade, /setup-browser-cookies, /setup-gbrain, /sync-gbrain, /autoplan, /pair-agent, /context-restore, /learn. Two CLI binaries also ship with the repo: gstack-model-benchmark and gstack-taste-update.

The big one for teams is /pair-agent. It coordinates Claude Code, Codex, and Hermes against the same task. Garry sits on top of this layer when he runs many sessions at once.

"I regularly run 10-15 parallel sprints. That's the practical max right now." (gstack README)

What Garry Says, in His Own Words

Three quotes worth keeping near the workflow because they shape how to use it.

On the philosophy: "A single builder with the right tooling can move faster than a traditional team." (gstack README)

On parallelism: "gstack is powerful with one sprint. It is transformative with ten running at once." (gstack README)

On the cost: "I sleep, like, four hours a night right now. I have cyber psychosis." (SXSW with Bill Gurley, March 2026)

That last one matters. Throughput at this rate is not the average user's normal day. Treat the headline numbers as a ceiling, not a floor.

The Reception Was Split

The launch trended on Product Hunt and pulled 33k stars in week one (now 97k+). Garry's tweet hit 849k views. A CTO friend called it god mode. (TechCrunch)

The pushback was loud too. Mo Bitar shipped a critique calling gstack "a bunch of prompts in a text file." Sherveen Mashayekhi argued the visibility came from Tan's YC role, not from the artifact's merit. The Hacker News thread has the full debate.

The steelman of the critique is fair. gstack does not run novel infrastructure. Each skill is markdown plus a system prompt. You could write the same files yourself in a weekend.

The counter is also fair. Most people will not write the same files in a weekend. They will write three, ship none, and revert to free-form prompting. A finished, opinionated set, vetted by someone shipping 10K lines a week, is a different artifact than your own half-written prompts.

Both readings are correct. Pick the one your team needs to hear.

Five Lessons for Founders

  1. Roles beat prompts. Defining a persona with priorities and constraints produces more consistent output than asking one Claude to wear seven hats.
  2. Forcing functions before code. /office-hours exists because reframing the product is cheaper than rewriting it.
  3. AI slop is a real failure mode. A dedicated Senior Designer skill (/plan-design-review) exists specifically to catch it.
  4. Ship is a workflow, not a button. /ship and /land-and-deploy chain sync, test, audit, push, deploy, verify into one command.
  5. Parallelism multiplies leverage. Garry runs 10 to 15 sprints at once. Solo builders get a structured team. Team leads get a fleet.

gstack vs Default Claude Code

Default Claude Codegstack
Free-form promptingSlash commands with defined roles
Whatever code style emergesArchitecture and design enforced by /plan-eng-review and /plan-design-review
Manual testing/qa opens a real browser, fixes bugs, writes regression tests
Manual ship steps/ship runs sync, test, audit, push, PR as one chain
Memory resets each session/learn plus GBrain carries patterns across sessions
Single agent/pair-agent coordinates Claude Code, Codex, Hermes

Clone gstack in 5 Minutes

You need Claude Code, Git, Bun v1.0+, and Node.js if you are on Windows. A Claude Pro subscription ($20/mo) is required for Claude Code itself.

Run the canonical install:

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

Then follow the loop. Start with /office-hours to reframe the product. Move to /plan-ceo-review and /plan-eng-review. Build. Run /review. Run /qa. Ship with /ship. That is the loop Garry is using to move at his stated rate.

Where gstack Stops

gstack gives you the workflow team. Twenty-three roles, a Think to Ship loop, parallel sprints, GBrain memory. It does not give you a SaaS to ship. It assumes you already have a product in front of it.

That gap is where Build This Now fits. Build This Now is the production skeleton the workflow operates on: auth (Google OAuth, email plus OTP, JWT, RLS), payments (Stripe checkout, subscriptions, webhooks, customer portal), database (PostgreSQL via Supabase with RLS on every table), email, file storage, background jobs, design system, landing page, legal pages, edge functions.

Adopt gstack on a blank repo and you spend the first three weeks rebuilding auth and payments before any of the 23 commands does real product work. Start with Build This Now and the skeleton is already there. gstack runs on top of it from day one. The workflow team plus the codebase the team builds on, $79 one-time on the CodeKit tier.

FAQ

What is gstack? An open-source collection of opinionated Claude Code skills built by Garry Tan that turns Claude Code into a virtual engineering team with seven specialist roles.

How many tools does gstack ship? 23 core slash commands as of 2026-05-15, plus around 15 supporting power tools and 2 CLI binaries.

What license? MIT.

Who created it? Garry Tan, President and CEO of Y Combinator. Open-sourced March 12, 2026.

How do I install it? git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setup

Does it require Claude Pro? Yes. Claude Code requires a Claude Pro subscription ($20/mo) or Team plan.

What is the role-based architecture? Each slash command embodies a specialist persona (CEO, Designer, Eng Manager, Release Manager, Doc Engineer, QA Lead, Chief Security Officer) with its own priorities, constraints, and outputs. Work moves through Think to Plan to Build to Review to Test to Ship to Reflect.

The CEO of YC published his exact Claude Code stack. The 23 commands are listed. The criticism is real. Pick the parts that fit your team and leave the rest.

Posted by @speedy_devv

Continue in Agents

  • Principes de base de l'agent
    Cinq façons de construire des agents spécialisés dans le code Claude : Sous-agents de tâches, .claude/agents YAML, commandes slash personnalisées, personas CLAUDE.md, et invites de perspective.
  • L'ingénierie du harness agent
    Le harness, c'est toutes les couches autour de ton agent IA sauf le modèle lui-même. Découvre les cinq leviers de contrôle, le paradoxe des contraintes, et pourquoi le design du harness détermine les performances de l'agent bien plus que le modèle.
  • Patterns d'agents
    Orchestrateur, fan-out, chaîne de validation, routage par spécialiste, raffinement progressif, et watchdog. Six formes d'orchestration pour câbler des sub-agents Claude Code.
  • Meilleures pratiques des équipes d'agents
    Patterns éprouvés pour les équipes d'agents Claude Code. Prompts de création riches en contexte, tâches bien calibrées, propriété des fichiers, mode délégué, et correctifs v2.1.33-v2.1.45.
  • Contrôles des équipes d'agents
    Configure le mode délégué, les modes d'affichage, l'approbation des plans, les limites de fichiers et les règles CLAUDE.md pour que le chef d'équipe Claude Code coordonne au lieu de coder.
  • Templates de prompts pour les équipes d'agents
    Dix prompts d'équipes d'agents testés pour Claude Code. Revue de code parallèle, débogage, builds de fonctionnalités, décisions d'architecture et recherche de campagne. À coller et utiliser.

More from Handbook

  • Changelog Claude Code
    Notes de version par version pour Claude Code, du bêta v0.2 à mars 2026. Bare mode, relais de permissions Channels, corrections OAuth, et chaque breaking change.
  • Templates de prompts qui livrent du code
    Dix recettes de prompts qui livrent du code : scaffolding full-stack, APIs, schémas, tests, refactors, debugging, reviews et CI. Chacun avec les modes d'échec à éviter.
  • FAQ Claude Code
    Réponses directes sur la tarification de Claude Code, les dépenses journalières, le choix de modèle, la comparaison avec Cursor, les Skills, CLAUDE.md, les clés API, et ce que l'agent terminal peut vraiment faire.
  • Dépannage de Claude Code
    Cinq vérifications dans l'ordre pour résoudre les pannes de Claude Code : erreurs d'installation, clés API incorrectes, timeouts 503, réponses lentes, et problèmes de permissions. La solution exacte pour chaque cas.

Arrêtez de configurer. Commencez à construire.

Templates SaaS avec orchestration IA.

Patterns d'agents

Orchestrateur, fan-out, chaîne de validation, routage par spécialiste, raffinement progressif, et watchdog. Six formes d'orchestration pour câbler des sub-agents Claude Code.

Des agents qui ressemblent à des humains

Des patterns de personnalité pour les agents Claude Code : raisonner à voix haute, admettre l'incertitude, peser les compromis, poser des questions. Un bloc CLAUDE.md à coller.

On this page

What gstack Is
Why It Matters
The Role-Based Architecture
The 23 Tools
A Real Workflow End to End
The Power Tools Most Posts Skip
What Garry Says, in His Own Words
The Reception Was Split
Five Lessons for Founders
gstack vs Default Claude Code
Clone gstack in 5 Minutes
Where gstack Stops
FAQ

Arrêtez de configurer. Commencez à construire.

Templates SaaS avec orchestration IA.