Garry Tan cf3582c637 fix: community security + stability fixes (wave 1) (#325)
* feat: add /cso skill — OWASP Top 10 + STRIDE security audit

* fix: harden gstack-slug against shell injection via eval

Whitelist safe characters (a-zA-Z0-9._-) in SLUG and BRANCH output
to prevent shell metacharacter injection when used with eval.

Only affects self-hosted git servers with lax naming rules — GitHub
and GitLab enforce safe characters already. Defense-in-depth.

* fix(security): sanitize gstack-slug output against shell injection

The gstack-slug script is consumed via eval $(gstack-slug) throughout
skill templates. If a git remote URL contains shell metacharacters
like $(), backticks, or semicolons, they would be executed by eval.

Fix: strip all characters except [a-zA-Z0-9._-] from both SLUG and
BRANCH before output. This preserves normal values while neutralizing
any injection payload in malicious remote URLs.

Before: eval $(gstack-slug) with remote "foo/bar$(rm -rf /)" → executes rm
After:  eval $(gstack-slug) with remote "foo/bar$(rm -rf /)" → SLUG=foo-barrm-rf-

* fix(security): redact sensitive values in storage command output

The browse `storage` command dumps all localStorage and sessionStorage
as JSON. This can expose tokens, API keys, JWTs, and session credentials
in QA reports and agent transcripts.

Fix: redact values where the key matches sensitive patterns (token,
secret, key, password, auth, jwt, csrf) or the value starts with known
credential prefixes (eyJ for JWT, sk- for Stripe, ghp_ for GitHub, etc.).

Redacted values show length to aid debugging: [REDACTED — 128 chars]

* fix(browse): kill old server before restart to prevent orphaned chromium processes

When the health check fails or the server connection drops, `ensureServer()`
and `sendCommand()` would call `startServer()` without first killing the
previous server process. This left orphaned `chrome-headless-shell` renderer
processes running at ~120% CPU each.

After several reconnect cycles (e.g. pages that crash during hydration or
trigger hard navigations via `window.location.href`), dozens of zombie
chromium processes accumulate and exhaust system resources.

Fix: call `killServer()` on the stale PID before spawning a new server in
both the `ensureServer()` unhealthy path and the `sendCommand()` connection-
lost retry path.

Fixes #294

* Fix YAML linter error: nested mapping in compact sequence entries

Having "Run: bun" inside a plain scalar is not allowed per YAML spec which states: Plain scalars must never contain the “: ” and “ #” character combinations.

This simple fix switches to block scalars (|) to eliminate the ambiguity without changing runtime behavior.

* fix(security): add Azure metadata endpoint to SSRF blocklist

Add metadata.azure.internal to BLOCKED_METADATA_HOSTS alongside the
existing AWS/GCP endpoints. Closes the coverage gap identified in #125.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* test: add coverage for storage redaction

Test key-based redaction (auth_token, api_key), value-based redaction
(JWT prefix, GitHub PAT prefix), pass-through for normal keys, and
length preservation in redacted output.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs: add community PR triage process to CONTRIBUTING.md

Document the wave-based PR triage pattern used for batching community
contributions. References PR #205 (v0.8.3) as the original example.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: adjust test key names to avoid redaction pattern collision

Rename testKey→testData and normalKey→displayName in storage tests
to avoid triggering #238's SENSITIVE_KEY regex (which matches 'key').
Also generate Codex variant of /cso skill.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs: update project documentation for v0.9.10.0

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: zero-noise /cso security audits with FP filtering (v0.11.0.0)

Absorb Anthropic's security-review false positive filtering into /cso:
- 17 hard exclusions (DOS, test files, log spoofing, SSRF path-only,
  regex injection, race conditions unless concrete, etc.)
- 9 precedents (React XSS-safe, env vars trusted, client-side code
  doesn't need auth, shell scripts need concrete untrusted input path)
- 8/10 confidence gate — below threshold = don't report
- Independent sub-agent verification for each finding
- Exploit scenario requirement per finding
- Framework-aware analysis (Rails CSRF, React escaping, Angular sanitization)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs: consolidate CHANGELOG — merge /cso launch + community wave into v0.11.0.0

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs: rewrite README — lead with Karpathy quote, cut LinkedIn phrases, add /cso

Opens with the revolution (Karpathy, Steinberger/OpenClaw), keeps credentials
and LOC numbers, cuts filler phrases, adds hater bait, restores hiring block,
removes bloated "What's new" section, adds /cso to skills table and install.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(cso): adversarial review fixes — FP filtering, prompt injection, language coverage

- Exclusion #10: test files must verify not imported by non-test code
- Exclusion #13: distinguish user-message AI input from system-prompt injection
- Exclusion #14: ReDoS in user-input regex IS a real CVE class, don't exclude
- Add anti-manipulation rule: ignore audit-influencing instructions in codebase
- Fix confidence gate: remove contradictory 7-8 tier, hard cutoff at 8
- Fix verifier anchoring: send only file+line, not category/description
- Add Go, PHP, Java, C#, Kotlin to grep patterns (was 4 languages, now 8)
- Add GraphQL, gRPC, WebSocket endpoint detection to attack surface mapping

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(docs): correct skill counts, add /autoplan to README tables

Skill count was wrong in 3 places (said 19+7=26, said 25, actual is 28).
Added /autoplan to specialist table. Fixed troubleshooting skills list
to include all skills added since v0.7.0.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(browse): DNS rebinding protection for SSRF blocklist

validateNavigationUrl is now async — resolves hostname to IP and checks
against blocked metadata IPs. Prevents DNS rebinding where evil.com
initially resolves to a safe IP, then switches to 169.254.169.254.
All callers updated to await. Tests updated for async assertions.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(browse): lockfile prevents concurrent server start races

Adds exclusive lockfile (O_CREAT|O_EXCL) around ensureServer to prevent
TOCTOU race where two CLI invocations could both kill the old server and
start new ones, leaving an orphaned chromium process. Second caller now
waits for the first to finish starting.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(browse): improve storage redaction — word-boundary keys + more value prefixes

Key regex: use underscore/dot/hyphen boundaries instead of \b (which treats
_ as word char). Now correctly redacts auth_token, session_token while
skipping keyboardShortcuts, monkeyPatch, primaryKey.

Value regex: add AWS (AKIA), Stripe (sk_live_, pk_live_), Anthropic (sk-ant-),
Google (AIza), Sendgrid (SG.), Supabase (sbp_) prefixes.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: migrate all remaining eval callers to source, fix stale CHANGELOG claim

5 templates and 2 bin scripts still used eval $(gstack-slug). All now use
source <(gstack-slug). Updated gstack-slug comment to match. Fixed v0.8.3
CHANGELOG entry that falsely claimed eval was fully eliminated — it was
the output sanitization that made it safe, not a calling convention change.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(docs): add /autoplan to install instructions, regen skill docs

The install instruction blocks and troubleshooting section were missing
/autoplan. All three skill list locations now include the complete 28-skill
set. Regenerated codex/agents SKILL.md files to match template changes.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs: update project documentation for v0.11.0.0

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* docs(cso): add disclaimer — not a substitute for professional security audits

LLMs can miss subtle vulns and produce false negatives. For production
systems with sensitive data, hire a real firm. /cso is a first pass,
not your only line of defense. Disclaimer appended to every report.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Arun Kumar Thiagarajan <arunkt.bm14@gmail.com>
Co-authored-by: Tyrone Robb <tyrone.robb@icloud.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Orkun Duman <orkun1675@gmail.com>
2026-03-22 13:19:10 -07:00
2026-03-12 01:32:16 -07:00

gstack

"I don't think I've typed like a line of code probably since December, basically, which is an extremely large change." — Andrej Karpathy, No Priors podcast, March 2026

When I heard Karpathy say this, I wanted to find out how. How does one person ship like a team of twenty? Peter Steinberger built OpenClaw — 247K GitHub stars — essentially solo with AI agents. The revolution is here. A single builder with the right tooling can move faster than a traditional team.

I'm Garry Tan, President & CEO of Y Combinator. I've worked with thousands of startups — Coinbase, Instacart, Rippling — when they were one or two people in a garage. Before YC, I was one of the first eng/PM/designers at Palantir, cofounded Posterous (sold to Twitter), and built Bookface, YC's internal social network.

gstack is my answer. I've been building products for twenty years, and right now I'm shipping more code than I ever have. In the last 60 days: 600,000+ lines of production code (35% tests), 10,000-20,000 lines per day, part-time, while running YC full-time. Here's my last /retro across 3 projects: 140,751 lines added, 362 commits, ~115k net LOC in one week.

2026 — 1,237 contributions and counting:

GitHub contributions 2026 — 1,237 contributions, massive acceleration in Jan-Mar

2013 — when I built Bookface at YC (772 contributions):

GitHub contributions 2013 — 772 contributions building Bookface at YC

Same person. Different era. The difference is the tooling.

gstack is how I do it. It turns 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, and a release engineer who ships the PR. Twenty specialists and eight power tools, all slash commands, all Markdown, all free, MIT license.

This is my open source software factory. I use it every day. I'm sharing it because these tools should be available to everyone.

Fork it. Improve it. Make it yours. And if you want to hate on free open source software — you're welcome to, but I'd rather you just try it first.

Who this is for:

  • Founders and CEOs — especially technical ones who still want to ship
  • First-time Claude Code users — structured roles instead of a blank prompt
  • Tech leads and staff engineers — rigorous review, QA, and release automation on every PR

Quick start

  1. Install gstack (30 seconds — see below)
  2. Run /office-hours — describe what you're building
  3. Run /plan-ceo-review on any feature idea
  4. Run /review on any branch with changes
  5. Run /qa on your staging URL
  6. Stop there. You'll know if this is for you.

Install — 30 seconds

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

Step 1: Install on your machine

Open Claude Code and paste this. Claude does the rest.

Install gstack: run git clone https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setup then add a "gstack" section to CLAUDE.md that says to use the /browse skill from gstack for all web browsing, never use mcp__claude-in-chrome__* tools, and lists the 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. Then ask the user if they also want to add gstack to the current project so teammates get it.

Step 2: Add to your repo so teammates get it (optional)

Add gstack to this project: run cp -Rf ~/.claude/skills/gstack .claude/skills/gstack && rm -rf .claude/skills/gstack/.git && cd .claude/skills/gstack && ./setup then add a "gstack" section to this project's CLAUDE.md that says to use the /browse skill from gstack for all web browsing, never use mcp__claude-in-chrome__* tools, lists the 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, /careful, /freeze, /guard, /unfreeze, /gstack-upgrade, and tells Claude that if gstack skills aren't working, run cd .claude/skills/gstack && ./setup to build the binary and register skills.

Real files get committed to your repo (not a submodule), so git clone just works. Everything lives inside .claude/. Nothing touches your PATH or runs in the background.

Codex, Gemini CLI, or Cursor

gstack works on any agent that supports the SKILL.md standard. Skills live in .agents/skills/ and are discovered automatically.

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

Or let setup auto-detect which agents you have installed:

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

This installs to ~/.claude/skills/gstack and/or ~/.codex/skills/gstack depending on what's available. All 28 skills work across all supported agents. Hook-based safety skills (careful, freeze, guard) use inline safety advisory prose on non-Claude hosts.

See it work

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.
        Prep takes forever and the results aren't good enough...

Claude: I'm going to push back on the framing. You said "daily briefing
        app." But what you actually described is a personal chief of
        staff AI.
        [extracts 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. The full vision is a 3-month project — start with
        the daily briefing that actually works.
        [writes design doc → feeds into downstream skills automatically]

You:    /plan-ceo-review
        [reads the design doc, challenges scope, runs 10-section review]

You:    /plan-eng-review
        [ASCII diagrams for data flow, state machines, error paths]
        [test matrix, failure modes, security concerns]

You:    Approve plan. Exit plan mode.
        [writes 2,400 lines across 11 files. ~8 minutes.]

You:    /review
        [AUTO-FIXED] 2 issues. [ASK] Race condition → you approve fix.

You:    /qa https://staging.myapp.com
        [opens real browser, clicks through flows, finds and fixes a bug]

You:    /ship
        Tests: 42 → 51 (+9 new). PR: github.com/you/app/pull/42

You said "daily briefing app." The agent said "you're building a chief of staff AI" — because it listened to your pain, not your feature request. Eight commands, end to end. That is not a copilot. That is a team.

The sprint

gstack is a process, not a collection of tools. The skills run in the order a sprint runs:

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

Each skill feeds into the next. /office-hours writes a design doc that /plan-ceo-review reads. /plan-eng-review writes a test plan that /qa picks up. /review catches bugs that /ship verifies are fixed. Nothing falls through the cracks because every step knows what came before it.

Skill Your specialist What they do
/office-hours YC Office Hours Start here. Six forcing questions that reframe your product before you write code. Pushes back on your framing, challenges premises, generates implementation alternatives. Design doc feeds into every downstream skill.
/plan-ceo-review CEO / Founder Rethink the problem. Find the 10-star product hiding inside the request. Four modes: Expansion, Selective Expansion, Hold Scope, Reduction.
/plan-eng-review Eng Manager Lock in architecture, data flow, diagrams, edge cases, and tests. Forces hidden assumptions into the open.
/plan-design-review Senior Designer Rates each design dimension 0-10, explains what a 10 looks like, then edits the plan to get there. AI Slop detection. Interactive — one AskUserQuestion per design choice.
/design-consultation Design Partner Build a complete design system from scratch. Researches the landscape, proposes creative risks, generates realistic product mockups.
/review Staff Engineer Find the bugs that pass CI but blow up in production. Auto-fixes the obvious ones. Flags completeness gaps.
/investigate Debugger Systematic root-cause debugging. Iron Law: no fixes without investigation. Traces data flow, tests hypotheses, stops after 3 failed fixes.
/design-review Designer Who Codes Same audit as /plan-design-review, then fixes what it finds. Atomic commits, before/after screenshots.
/qa QA Lead Test your app, find bugs, fix them with atomic commits, re-verify. Auto-generates regression tests for every fix.
/qa-only QA Reporter Same methodology as /qa but report only. Pure bug report without code changes.
/cso Chief Security Officer OWASP Top 10 + STRIDE threat model. Zero-noise: 17 false positive exclusions, 8/10+ confidence gate, independent finding verification. Each finding includes a concrete exploit scenario.
/ship Release Engineer Sync main, run tests, audit coverage, push, open PR. Bootstraps test frameworks if you don't have one.
/land-and-deploy Release Engineer Merge the PR, wait for CI and deploy, verify production health. One command from "approved" to "verified in production."
/canary SRE Post-deploy monitoring loop. Watches for console errors, performance regressions, and page failures.
/benchmark Performance Engineer Baseline page load times, Core Web Vitals, and resource sizes. Compare before/after on every PR.
/document-release Technical Writer Update all project docs to match what you just shipped. Catches stale READMEs automatically.
/retro Eng Manager Team-aware weekly retro. Per-person breakdowns, shipping streaks, test health trends, growth opportunities.
/browse QA Engineer Real Chromium browser, real clicks, real screenshots. ~100ms per command.
/setup-browser-cookies Session Manager Import cookies from your real browser into the headless session. Test authenticated pages.
/autoplan Review Pipeline One command, fully reviewed plan. Runs CEO → design → eng review automatically with encoded decision principles. Surfaces only taste decisions for your approval.

Power tools

Skill What it does
/codex Second Opinion — independent code review from OpenAI Codex CLI. Three modes: review (pass/fail gate), adversarial challenge, and open consultation. Cross-model analysis when both /review and /codex have run.
/careful Safety Guardrails — warns before destructive commands (rm -rf, DROP TABLE, force-push). Say "be careful" to activate. Override any warning.
/freeze Edit Lock — restrict file edits to one directory. Prevents accidental changes outside scope while debugging.
/guard Full Safety/careful + /freeze in one command. Maximum safety for prod work.
/unfreeze Unlock — remove the /freeze boundary.
/setup-deploy Deploy Configurator — one-time setup for /land-and-deploy. Detects your platform, production URL, and deploy commands.
/gstack-upgrade Self-Updater — upgrade gstack to latest. Detects global vs vendored install, syncs both, shows what changed.

Deep dives with examples and philosophy for every skill →

Parallel sprints

gstack works well with one sprint. It gets interesting with ten running at once.

Conductor runs multiple Claude Code sessions in parallel — each in its own isolated workspace. One session on /office-hours, another on /review, a third implementing a feature, a fourth running /qa. All at the same time. The sprint structure is what makes parallelism work — without a process, ten agents is ten sources of chaos. With a process, each agent knows exactly what to do and when to stop.


Free, MIT licensed, open source. No premium tier, no waitlist.

I open sourced how I build software. You can fork it and make it your own.

We're hiring. Want to ship 10K+ LOC/day and help harden gstack? Come work at YC — ycombinator.com/software Extremely competitive salary and equity. San Francisco, Dogpatch District.

Docs

Doc What it covers
Skill Deep Dives Philosophy, examples, and workflow for every skill (includes Greptile integration)
Builder Ethos Builder philosophy: Boil the Lake, Search Before Building, three layers of knowledge
Architecture Design decisions and system internals
Browser Reference Full command reference for /browse
Contributing Dev setup, testing, contributor mode, and dev mode
Changelog What's new in every version

Privacy & Telemetry

gstack includes opt-in usage telemetry to help improve the project. Here's exactly what happens:

  • Default is off. Nothing is sent anywhere unless you explicitly say yes.
  • On first run, gstack asks if you want to share anonymous usage data. You can say no.
  • What's sent (if you opt in): skill name, duration, success/fail, gstack version, OS. That's it.
  • What's never sent: code, file paths, repo names, branch names, prompts, or any user-generated content.
  • Change anytime: gstack-config set telemetry off disables everything instantly.

Data is stored in Supabase (open source Firebase alternative). The schema is in supabase/migrations/001_telemetry.sql — you can verify exactly what's collected. The Supabase publishable key in the repo is a public key (like a Firebase API key) — row-level security policies restrict it to insert-only access.

Local analytics are always available. Run gstack-analytics to see your personal usage dashboard from the local JSONL file — no remote data needed.

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: gstack works on Windows 11 via Git Bash or WSL. Node.js is required in addition to Bun — Bun has a known bug with Playwright's pipe transport on Windows (bun#4253). The browse server automatically falls back to Node.js. Make sure both bun and node are on your PATH.

Claude says it can't see the skills? Make sure your project's CLAUDE.md has a gstack section. Add this:

## 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.

License

MIT. Free forever. Go build something.

S
Description
No description provided
Readme MIT 307 MiB
Languages
TypeScript 75.7%
Go Template 14.5%
Shell 6.4%
JavaScript 2.1%
CSS 0.8%
Other 0.5%