codexlisted

# Codex Health and Parity Check Use this skill when you want a fast answer to: - Is Codex using ChatGPT auth instead of API billing? - Are curated Claude skills installed natively in Codex? - Does `~/.codex/AGENTS.md` point at canonical `CLAUDE.md`? - Are native skills and the MCP profile synchronized? - What exactly is out of sync if parity drifted? All commands below use `~/.claude/scripts/`, the installed (symlinked) location — not a repo-relative path — so this works regardless of which repo's `install.sh` put the scripts there. ## Default mode Run the repo health script: ```bash bash ~/.claude/scripts/codex-check.sh ``` Then summarize the result in 3 sections: 1. **Auth** — chatgpt vs apikey, and whether billing risk exists 2. **Surface** — native skills, canonical instructions, MCP profile 3. **Action** — exact next command if repair is needed ## `--profile` mode Install one or more explicit Codex skill packs, then re-run health. Packs are additive to the engineering default. Available pack names vary by installed overlay — run `--profiles` mode first to list what's actually available: ```bash bash ~/.claude/scripts/setup-codex-skill-profile.sh --profile <pack[,pack...]> bash ~/.claude/scripts/codex-check.sh ``` Accept comma-separated packs and the special profiles `default` and `all`. Example: ```bash # Claude Code /bs:codex --profile <pack1>,<pack2> # Local Codex or Terminal: choose one (permission is required for ~/.agents/skills) bash ~/.claude/script