gstack is Garry Tan's set of slash-command skills, written in Markdown. They turn Claude Code and other coding agents into a virtual engineering team that works through a sprint: think, plan, build, review, test, ship and reflect. It is written for founders, first-time Claude Code users and tech leads who want structured roles and review gates in place of a blank prompt.
Why read it
The long skills keep the main file as a short spine. Every step that has detailed rules gets a STOP block that names the section file to read in full first. A self-check at the end makes the agent list which sections it read and redo any step it did from memory. This keeps the main file small and stops the agent from working from a remembered version of the rules.
Each skill starts by running a helper script that prints KEY: value status lines, and the rules that follow branch on those lines. The skill also spells out a degraded mode with safe defaults for when the script is missing, so a stale install still finishes the user's task.
The skill follows embedded instruction blocks only when they come from the command it just ran and carry the matching session ID. PR text, hunk text and web answers are labelled as untrusted data. Preference changes are written only when the user types them in chat. Together these rules defend against prompt injection from tool output.
The review skill makes the agent quote file:line evidence before it reports a finding. A finding it cannot quote gets a confidence of 4 to 5. A table then maps each score band to a display rule: shown normally, shown with a caveat, or kept to the appendix. This cuts false positives without hiding anything outright.
Skills report completion with one of four statuses: DONE, DONE_WITH_CONCERNS, BLOCKED or NEEDS_CONTEXT. They escalate after three failed attempts. A missing or partial check must be reported as missing coverage, never as a pass. An empty learnings review has to be stated out loud rather than skipped silently.
The review skill sorts each finding into AUTO-FIX or ASK. It applies mechanical fixes directly and batches the ASK items into one question with lettered options. Exactly one option carries a (recommended) label. Some findings, such as edits that weaken tests or other gates, are always ASK, and a headless session defaults to leaving them open.
A top-level router skill maps request patterns and voice-transcription variants to specific skills. A user setting (off, core or all) controls how freely the agent invokes skills without being asked. The router also routes only to skills that are actually installed.
The /ship skill keeps a private Markdown record of the run, with fixed headings for releases, decisions, reviews, checks and next steps. Repairs are inserted as explicit step ranges into a saved work list, so a run that pauses or loops can pick up where it left off without dropping work that is still pending.
The drawing appears once a model has read every skill here.
Every one is a skill. Started by: You or the agent.
| SKILL | DESCRIPTION | LINKS |
|---|---|---|
| autoplan | Auto-review pipeline — reads the full CEO, design, eng, and DX review skills from disk and runs them sequentially with auto-decisions using 6 decision principles. (gstack) | 10 |
| benchmark | Web performance regression detection. (gstack) | 2 |
| benchmark-models | Cross-model benchmark for gstack skills. (gstack) | 2 |
| browse | Drive a real browser through Aside: open a page, read it, click through a flow, take screenshots, check console errors. (gstack) | 12 |
| canary | Post-deploy canary monitoring. (gstack) | 11 |
| captain | Run a crew of coding agents from one controller (also /crew): lanes send READY blocks and never merge; the controller merges through trains, keeps the digest, relays approvals, fixes flakes daily. (gstack) | 3 |
| careful | Safety guardrails for destructive commands. (gstack) | 1 |
| codex | OpenAI Codex CLI wrapper — three modes. (gstack) | 12 |
| context-restore | Restore working context saved earlier by /context-save. (gstack) | 2 |
| context-save | Save working context. (gstack) | 5 |
| cso | Security audit: supported static findings; qualified profiles add reproduction and repair candidates. (gstack) | 1 |
| design-consultation | Design consultation: understands your product, researches the landscape, proposes a complete design system (aesthetic, typography, color, layout, spacing, motion), and generates font+color preview... (gstack) | 5 |
| design-html | Design finalization: generates production-quality Pretext-native HTML/CSS. (gstack) | 7 |
| design-motion | Design motion: animate an approved design into a seamless MP4 and GIF loop. (gstack) | 6 |
| design-review | Designer's eye QA: finds visual inconsistency, spacing issues, hierarchy problems, AI slop patterns, and slow interactions — then fixes them. (gstack) | 7 |
| design-shotgun | Design shotgun: generate multiple AI design variants, open a comparison board, collect structured feedback, and iterate. (gstack) | 6 |
| deslop-shared-libs | Find worthwhile shared-code extractions in recent work. (gstack) | 2 |
| devex-review | Live developer experience audit. (gstack) | 12 |
| diagram | Turn an English description (or mermaid source) into a diagram triplet: the source, an editable .excalidraw file you can open on excalidraw.com, and rendered SVG + PNG. (gstack) | 1 |
| document-generate | Generate missing documentation from scratch for a feature, module, or entire project. (gstack) | 4 |
| document-release | Release documentation audit. (gstack) | 2 |
| eval-plan | Preregister a paid eval (bars, arms, decoys, stop rules), dry-run it at zero cost, then run a three-item priced pilot whose spend table gates approval. (gstack) | 3 |
| freeze | Restrict file edits to a specific directory for the session. (gstack) | 3 |
| gstack | Router for the gstack skill suite. (gstack) | 56 |
| gstack-upgrade | Upgrade gstack to the latest version. | 1 |
| guard | Full safety mode: destructive command warnings + directory-scoped edits. (gstack) | 4 |
| health | Code quality dashboard. (gstack) | 5 |
| investigate | Systematic debugging with root cause investigation. (gstack) | 3 |
| ios-clean | Remove the DebugBridge SPM package and all #if DEBUG wiring from an iOS app. (gstack) | 6 |
| ios-design-review | Visual design audit for iOS apps on real hardware. (gstack) | 5 |
| ios-fix | Autonomous iOS bug fixer. (gstack) | 8 |
| ios-qa | Live-device iOS QA for SwiftUI apps. (gstack) | 10 |
| ios-sync | Regenerate the iOS debug bridge against the latest upstream gstack templates. (gstack) | 1 |
| land-and-deploy | Land and deploy workflow. (gstack) | 16 |
| landing-report | Read-only queue dashboard for workspace-aware ship. (gstack) | 5 |
| learn | Manage project learnings. | 7 |
| make-pdf | Turn any markdown file into a publication-quality PDF. (gstack) | 1 |
| modernize | Bring an aging pull request onto its current base and decide whether it should still exist: MODERNIZE, REWORK or CLOSE, with range-diff proof before any push. (gstack) | 5 |
| office-hours | YC Office Hours — two modes. (gstack) | 9 |
| open-gstack-browser | Launch GStack Browser — AI-controlled Chromium with the sidebar extension baked in. | 4 |
| pair-agent | Pair a remote AI agent with your browser. (gstack) | 3 |
| plan-ceo-review | CEO/founder-mode plan review. (gstack) | 8 |
| plan-design-review | Designer's eye plan review — interactive, like CEO and Eng review. (gstack) | 14 |
| plan-devex-review | Interactive developer experience plan review. (gstack) | 12 |
| plan-eng-review | Eng manager-mode plan review. (gstack) | 9 |
| plan-tune | Self-tuning question sensitivity + developer psychographic for gstack (v1: observational). (gstack) | 5 |
| qa | Test the site, API or CLI and fix the bugs it finds. (gstack) | 11 |
| qa-only | Report browser/API/CLI/job/worker/webhook bugs. (gstack) | 5 |
| retro | Weekly engineering retrospective. (gstack) | 9 |
| review | Review this PR's diff. (gstack) | 12 |
| scrape | Pull data from a web page through the Aside browser — your real, already signed-in sessions. (gstack) | 4 |
| setup-browser-cookies | Import cookies from your real Chromium browser into the headless browse session. (gstack) | 1 |
| setup-deploy | Configure deployment settings for /land-and-deploy. | 2 |
| setup-gbrain | Set up gbrain for this coding agent: install the CLI, initialize a local PGLite or Supabase brain, register MCP, capture per-remote trust policy. (gstack) | 4 |
| ship | Ship it: merge the base, run tests, review the diff, bump VERSION and CHANGELOG, push and create a PR. (gstack) | 14 |
| skillify | Codify the most recent successful /scrape flow into a permanent browser-skill on disk. (gstack) | 5 |
| spec | Turn vague intent into a precise, executable spec in five phases. (gstack) | 11 |
| sync-gbrain | Keep gbrain current with this repo's code and refresh agent search guidance in CLAUDE.md. (gstack) | 5 |
| test-audit | Find low-value or duplicate tests and the test-only code they keep alive. (gstack) | 8 |
| ubicloud | Run a heavy suite, build or eval on an ephemeral 16-vCPU Ubicloud VM: TTL-named VMs, a reaper, quota preflight and a hang watchdog. (gstack) | 3 |
| unfreeze | Clear the freeze boundary set by /freeze, allowing edits to all directories again. (gstack) | 2 |
| verify | Build loop while coding: run the repo's checks and the tests /ship would pick, fix, repeat until green. (gstack) | 2 |
| write-skill | Write a new agent skill or improve one without bloat: the edit ladder, levels, drift and size checks, full diets. (gstack) | 2 |
From garrytan/gstack at 3f79d81, 2026-11-01. Skills read by DeepSeek V4.1 Flash on 2026-11-04. Summary and techniques by Claude Opus 5.5 on 2026-11-06.