Brainstorm and Shape Ideas Before Code8 skills
YC-style product interrogation and specification that forces thinking into a design doc.
Engineering Blueprint
Safety checkedRun AI-Assisted Development Workflows End to End
garrytanMay 2026
A suite of 23+ specialized skills that route Claude Code through structured workflows for planning, review, testing, and shipping. Covers ideation through production.
Ideation & Problem FramingStrategic Planning & ScopeArchitecture & Technical ReviewDesign & Visual SystemsDevelopment & CodingCode Quality & Pre-Landing ReviewTesting & QADebugging & TroubleshootingRelease & Deployment
Install Command
Run this command to deploy the blueprint to your environment.
What problem does this solve?
Ships your full development lifecycle from thinking through reflecting. Each stage (think, plan, build, review, test, ship, reflect) has a dedicated specialist and carries output forward to the next, so context flows instead of getting lost between steps. This removes the tax of coordinating multiple AI tools and prevents rework from restarting cold at each stage.
How does it work?
After setup, you invoke skills sequentially as you work. 1. /office-hours or /spec shape the work (problem statement, design doc, alternatives). 2. /plan-ceo-review, /plan-eng-review, and other review skills lock in decisions before code. 3. Build your code (Cursor, Claude Code, or text editor). 4. /review audits the diff; /qa tests the site; /investigate debugs broken things. 5. /ship automates the merge, tests, version bump, changelog, and PR. 6. /retro captures learnings. Each skill reads prior outputs, so every downstream decision has full context.
Included Skill Groups
Plan Scope and Strategy With CEO and Engineering Review14 skills
Rethink scope ambition and lock down architecture before writing code.
Write Code in Your Editor or Claude Code10 skills
Implement the locked plan using your preferred coding environment.
Audit Code Quality and Find Bugs Before Landing15 skills
Specialist-driven pre-landing review that catches what CI doesn't (architecture, security, edge cases).
Test Code and Find Bugs in Live Deployments13 skills
Automated browser-based QA that tests real user flows and generates regression tests.
Debug Broken Behavior With Systematic Root-Cause Investigation11 skills
Structured troubleshooting that traces data flow and tests hypotheses instead of guessing.
Ship Code End to End: Test, Bump Version, Create PR, Deploy9 skills
Fully automated release workflow from merge base to production verification.
Capture Learnings and Run Retrospectives9 skills
Persistent memory and process reflection that compounds across sessions.
What's the biggest win?
One person ships like a team of specialists because each stage knows what came before. Your context never resets, decisions propagate forward, and quality gates run automatically.
What should I know technically?
Requires Claude Code (Anthropic's in-IDE agent) as the primary entry point. Skills auto-detect and support 10+ other agents (Cursor, Codex CLI, OpenCode, Factory Droid, Slate, Kiro, Hermes, GBrain mod). Core runtime is Bash, TypeScript, and Bun. Optional integrations: GitHub/GitLab CLI for PR creation, Playwright for browser QA, Supabase/PGLite for GBrain (persistent memory across sessions). The /browse skill requires a local Chromium instance built on first use. WIP checkpoint mode auto-commits; team mode syncs gstack across developers in the same repo.
What should I watch out for?
gstack is a workflow orchestrator for advanced users who run Claude Code regularly. It's not for occasional usage or as a copilot replacement. The most powerful payoff comes from running the full 7-stage sprint; isolated skill invocation works but doesn't accumulate context. Requires Claude Code, Git, and Bun runtime. Early versions may have rough edges. Report issues to the repo.
Key Features
End-to-End Workflow
7-stage sprint (think, plan, build, review, test, ship, reflect) where each skill feeds into the next.
Specialist Roles
23+ skills with defined personas (CEO, Designer, QA Lead, Security Officer) that carry specialized judgment into each phase.
Context Carryover
Each skill reads outputs from the prior stage, so decisions propagate automatically without rewriting briefings.
Quality Gates
Automatic code review, test coverage audit, plan completion verification, and pre-landing adversarial review before shipping.
Production Shipping
Automates version bumps, CHANGELOG generation, PR creation, and post-deploy health checks in one `/ship` command.
Tools in this Blueprint
About This Blueprint
- License
- MIT
- Industry
- Technology
- Skills
- 2 workflows, 15 sub-skills, 6 standalone