What problem does it solve?
It removes the repetitive back-and-forth of running multiple review passes by automatically reading your CEO, design, and engineering review skills from disk and executing them in order, so you get a fully reviewed plan without answering dozens of intermediate prompts.
Core Features & Use Cases
- Manual trigger only pipeline: Runs only when the user explicitly invokes /autoplan, to avoid unexpected behavior.
- Sequential full review gauntlet: Executes CEO → Design (when UI scope is detected) → Eng reviews, enforcing the required order and producing all required outputs.
- Auto-decided intermediate questions: Applies six decision principles to choose between options, while surfacing taste/borderline decisions at a final approval gate.
- Dual-voice strategy and design review: Runs an independent CEO/design subagent plus a Codex voice (when available) to improve coverage and catch strategic or UX blind spots.
- Final “approval gate” with surfaced taste decisions: Keeps mechanical decisions automatic while flagging where reasonable people could disagree.
Quick Start
Invoke /autoplan to have gstack run the full CEO → (Design if applicable) → Eng review pipeline and output a fully reviewed plan.