agentleFS
Sign inStart free
s977043s977043/river-review/.claude/skills/ask-codex/SKILL.md

ask-codex

Skill7 starsChanged 5 months agoMITView on GitHub

Get a second opinion or independent review from Codex (OpenAI) via the local Codex CLI. Use when the user wants Codex to review a design memo, PR, diff, or decision, or to cross-check an approach from another model. Also trigger on "Codexに聞いて", "Codexと相談", "Codexにレビュー", "Codexの意見", "別のモデルで確認".

  • Reads credentials

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Handle authentication failure: If the output contains 401 Unauthorized / Missing bearer / Reconnecting..., Codex is not authenticated.

    line 48
  2. Very specific:

    Never paste secrets or .env contents into the Codex prompt.

    line 67

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Purpose

Delegate a focused question or review to Codex (OpenAI's CLI) and relay its answer back to the user as an independent, attributed second opinion.

This skill does not replace your own judgment — it gathers Codex's view so the user can compare it against yours (and against other reviewers such as Antigravity).

How Codex is invoked

The repository ships a project-local Codex config (.codex/config.toml) and a wrapper script. Always call Codex through the script so the project config and working directory are correct:

npm run codex:exec -- "<prompt>"

This runs codex exec non-interactively (CODEX_HOME=.codex, -C <repo root>) and prints Codex's answer to stdout. Codex runs in the repo, so it can read files itself — reference paths in the prompt instead of pasting large files.

Steps

  1. Frame the prompt. Turn the user's request into a single self-contained prompt for Codex. Include:
    • the role ("You are a senior reviewer of River Review …"),
    • the file paths Codex should read (e.g. docs/development/1255-…md, runners/core/review-runner.mjs),
    • the specific questions to answer, numbered,
    • a request to be concise and to state a clear conclusion.

  2. Run Codex with a generous timeout (design reviews can take minutes). Set the timeout on the Bash tool call itself (e.g. timeout: 600000 ms) — do not prefix the command with timeout 600: macOS ships no GNU timeout binary, so that prefix fails with command not found.

    npm run codex:exec -- "<prompt>"
  3. Handle authentication failure. If the output contains 401 Unauthorized / Missing bearer / Reconnecting..., Codex is not authenticated. Do not retry blindly. Tell the user to authenticate themselves in this session (these are interactive / secret-bearing and you cannot run them for them):
    • ! CODEX_HOME="$(git rev-parse --show-toplevel)/.codex" codex login (the CODEX_HOME prefix is required — a bare codex login writes credentials to ~/.codex and the project-local config keeps failing), or
    • set OPENAI_API_KEY in the environment (warn that pasting a key into chat leaves it in history; prefer env/keychain). Then offer to re-run once authenticated.
  4. Relay faithfully. Quote or summarize Codex's answer and attribute it to Codex ("Codex の見解:"). Keep Codex's conclusion distinct from your own. If you disagree, say so separately — do not silently overwrite its opinion.
  5. Persist when useful. For design decisions, offer to record Codex's verdict in the relevant memo / issue / PR so it is not lost.

Guardrails

  • Never paste secrets or .env contents into the Codex prompt.
  • Codex executes in workspace-write sandbox; treat its file edits as proposals to review, not as automatically trusted changes.
  • This skill is for advisory review. Decisions and commits stay with the user and with you.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free