Step-by-step installation and setup of gstack on Claude Code, Codex, Cursor, and OpenClaw—covering how it works, how to verify it's active, and how to troubleshoot common issues. Includes safety guardrails: /careful, /freeze, /guard.
/ship syncs main, runs tests, audits coverage, pushes, opens PR. /land-and-deploy merges, waits for CI, deploys, verifies production. /canary monitors post-deploy. /benchmark tracks performance baseline.
/document-release updates all docs to match what shipped. /retro runs weekly engineering retrospectives with per-person breakdowns. /learn manages patterns, pitfalls, and preferences for your codebase.
/careful warns on destructive commands. /freeze locks edits to one directory. /guard combines both. /investigate systematically debugs with auto-freeze. Conductor coordinates 10-15 parallel Claude Code sessions.
office-hours is gstack's entry point skill. It asks six forcing questions that reframe your product before you write code. This guide shows how YC-style product interrogation turns vague ideas into clear requirements.
plan-ceo-review finds the 10-star product hiding inside your request. Four scope modes, strategic challenge, and the /autoplan pipeline for automatic review routing.
Three plan review skills for different audiences: /plan-eng-review for architecture, /plan-devex-review for developer experience, /plan-design-review for UI/UX. Learn which review to use when.
The design pipeline turns ideas into production HTML: /design-consultation builds your design system, /design-shotgun generates variants, /design-html produces shippable code with Pretext computed layout.
/browse gives the AI agent eyes. Real Chromium browser, actual clicks, screenshots. ~100ms per command. Includes sidebar agent with auto model routing and handoff/resume for CAPTCHA handling.
/pair-agent shares browser between AI agents. Cross-agent coordination with scoped tokens, tab isolation, and cookie import for authenticated testing.