What problem does it solve?
It helps you get higher-confidence code outcomes by running a structured OpenAI Codex workflow for independent review, adversarial challenge, and ongoing consult-style follow-ups.
Core Features & Use Cases
- Three Codex modes (review, challenge, consult): perform independent diff-based review with pass/fail gating, run an adversarial mode to try to break your code, or consult Codex with session continuity for iterative follow-ups.
- Prompting for “codex” intent with routing-friendly triggers: use natural-language or voice trigger variants that map to the Codex wrapper behavior for predictable activation.
- Execution in a plan-safe, tool-scoped workflow: supports planning-phase guidance while still enabling AskUserQuestion as the workflow entry/exit gate for interactive decisions.
Use case: Before merging a PR, ask for “codex review” to validate your diff, then ask for “codex challenge” to surface edge cases and failure modes you missed, and finally “consult codex” to iterate on fixes with continuity.
Quick Start
Use the gstack-codex skill when you want codex review, codex challenge, or a second opinion about your code by asking for “codex review” or “codex challenge” (or using the voice aliases “code x” / “code ex” / “get another opinion”).