What problem does it solve?
Running OpenAI's Codex CLI non-interactively from scripts or other agents is prone to common footguns like infinite hangs from unclosed stdin, noisy output mixing final answers with progress logs, and unexpected behavior from heavy local config files, making automation unreliable without explicit guidance.
Core Features & Use Cases
- Stdin hang prevention: Clear guidance for closing stdin correctly across bash and PowerShell to avoid blocked runs.
- Clean output capture: Isolates the final agent response from stderr logs and reasoning streams for context-lean downstream use.
- Configurable safety and control: Supports sandbox mode selection, model and reasoning effort overrides, and clean config flags for reproducible unattended runs in CI or cross-agent workflows.
- Use case example: A developer can use this skill to run a headless code review of their repository from a CI pipeline and save only the final review summary to a file for later processing.
Quick Start
Use the codex-cli skill to run a headless analysis of your current repository and save only the final review summary to a file named repo-review.md.