
from claude-code-skills20
Coordinate calls to the Codex CLI: collect minimal context, run codex exec/resume, and present the model's independent analysis or review to the user.
This skill acts as a coordinator for invoking the Codex CLI to obtain an independent model opinion. It transforms user intent into Codex CLI commands, gathers only the necessary context (git diffs, files, or conversation snippets), runs codex exec or codex exec resume, and parses the JSONL output to extract session/thread IDs and results. It enforces safe defaults (read-only sandboxing) and filters sensitive files from context.
Trigger when the user explicitly asks to consult Codex for analysis, review, or a second opinion on code, design, or proposals (phrases like ask Codex or let Codex review). Do NOT trigger for installation or troubleshooting questions about the Codex CLI itself. Use this for longer-running exploratory tasks that may execute in background mode.
Agents that can run shell commands and manage background tasks (CLI-capable agents). Designed to integrate with systems that can run the Codex CLI and parse JSONL outputs.
This skill has not been reviewed by our automated audit pipeline yet.