gstack

中文 | English

Summary

This source packages Garry Tan's Claude Code environment as a multi-role engineering office: product review, design critique, QA, security, release, and retrospective flows are all exposed as reusable commands and team bootstrap steps.

Source

• Raw file: raw/garrytangstack Use Garry Tan's exact Claude Code setup 23 opinionated tools that serve as CEO, Designer, Eng Manager, Release Manager, Doc Engineer, and QA.md
• Translated raw file: raw/garrytangstack Use Garry Tan's exact Claude Code setup 23 opinionated tools that serve as CEO, Designer, Eng Manager, Release Manager, Doc Engineer, and QA.zh.md
• Original URL: https://github.com/garrytan/gstack
• Ingest date: 2026-04-15

Key Contributions

• Frames a single agent environment as a reusable “virtual engineering team” with named specialist roles.
• Makes team distribution explicit: global install for each developer plus repo bootstrap so project defaults travel with the codebase.
• Shows how one methodology can be routed through Claude Code, OpenClaw, and other AI agent environments.
• Pushes browser QA, security review, release, and retros into slash-command surfaces instead of leaving them as ad hoc reminders.

Practical Implications

• Agent-first repos increasingly need explicit bootstrap and upgrade paths, not just good prompts.
• Named roles can help a repo encode expected review perspectives even when the underlying model is the same.
• The more throughput a team wants, the more value there is in making “support” work like QA and release automation agent-native too.

Related Pages

• Agent-First Repositories
• Agent Harness Design
• Codex Operating Practices