ask-codex — Claude Code skill: query GPT-5, how to install
ask-codex is a Claude Code skill that sends a question about the current repo to OpenAI Codex (GPT-5) and returns the answer without a subagent. For one-off questions, a second opinion, and bug searches.
MIT ★ 195 updated 2026-08-10
/plugin marketplace add Sateezg/codex-bridge
/plugin install codex-bridge@codex-bridgeWhat ask-codex does
The skill calls codex-run, which sends a question to Codex, reads the repo, and returns the final answer on stdout. Codex runs in read-only mode by default and cannot modify files.
Each question must be self-contained: Codex has no knowledge of the current conversation. Name the files or directories to look at, state what you already tried, specify the answer format (yes/no, line number, brief cause). Example: `codex-run -C /path/to/repo "In src/auth/session.ts, refreshToken() occasionally returns a token that fails validation immediately. I suspect a clock-skew issue in the exp comparison. Read that file plus src/auth/jwt.ts and tell me the single most likely root cause and the exact line. Be concise."`
Pass -s workspace-write to let Codex modify files, -r to continue the previous session for a follow-up question, --timeout 1800 to raise the wait limit for large repos. Each call spends the user's ChatGPT plan quota: one question, one call. Codex may cite files or lines that do not exist: verify any specific claim before acting on it.
What it helps with
- get a second opinion on an architectural decision
- find the root cause of a bug in an unfamiliar part of the codebase
- ask GPT-5 a one-off question about the repo without starting a subagent
Example prompts for the agent
- Ask Codex why
refreshToken()returns an invalid token - Ask GPT which file handles session caching
- Get a second opinion on the auth module design
How to install ask-codex
1Plugin marketplace
/plugin marketplace add Sateezg/codex-bridge/plugin install codex-bridge@codex-bridgeThis installs the whole codex-bridge plugin with all its skills
2Manually: put the files in place
~/.claude/skills/ask-codex/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/ask-codex/
Manually: put the files in place
.agents/skills/ask-codex/SKILL.mdFor all projects use ~/.agents/skills/
Manually: put the files in place
.cursor/skills/ask-codex/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install Sateezg/codex-bridge skills/ask-codex2Manually: put the files in place
.github/skills/ask-codex/SKILL.mdPlatform CLI
gemini skills install https://github.com/Sateezg/codex-bridge --path skills/ask-codexManually: put the files in place
.devin/skills/ask-codex/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about ask-codex
How to install ask-codex in Claude Code?
Run /plugin marketplace add Sateezg/codex-bridge, then /plugin install codex-bridge@codex-bridge. For Codex, use npx skills add Sateezg/codex-bridge --skill ask-codex -a codex.
What does ask-codex do?
The skill sends a question to OpenAI Codex (GPT-5) about the current repo and returns the answer without creating a subagent. Codex reads code by default but does not modify it.
Can I ask multiple questions in a row?
Yes, use the -r flag to continue the previous session: codex-run -C /path/to/repo -r "Now show me the minimal patch". This keeps Codex's context and avoids re-reading the repo.
Source text
SKILL.md · MIT
---
description: Ask OpenAI Codex (GPT-5) a one-off question about the current repo and get its answer back, without spinning up a subagent. Use when the user says "ask codex", "what does GPT think", "get a second opinion on this", or when you want an independent read on a design call, a tricky bug, or an unfamiliar area of the codebase.
---
# Ask Codex a question
For a single question with a single answer. Anything multi-step or long-running
belongs to a subagent (`codex-second-opinion`, `codex-reviewer`, `codex-debugger`).
```bash
codex-run -C <repo-dir> "<self-contained question>"
```
`codex-run` prints **only Codex's final answer** on stdout (it captures the final
message via `codex exec -o` rather than dumping the whole event log). It defaults
to a read-only sandbox, so Codex can read the repo but cannot change anything.
## Writing the question
Codex has zero knowledge of this conversation. Every question must stand alone:
- Name the files or directories it should look at.
- State what you already know or already tried.
- Say what shape of answer you want ("name the single most likely cause and the
line to look at", "answer yes/no then justify in three sentences").
```bash
codex-run -C /path/to/repo \
"In src/auth/session.ts, refreshToken() occasionally returns a token that fails validation immediately. I suspect a clock-skew issue in the exp comparison. Read that file plus src/auth/jwt.ts and tell me the single most likely root cause and the exact line. Be concise."
```
## Useful options
```
-s workspace-write let Codex modify files (only when the user asked for that)
-m gpt-5-codex model override
-r continue the previous codex session (follow-up question)
--timeout 1800 raise the 900s default for a big repo sweep
--raw also print Codex's full event log to stderr, for debugging
```
Use `-r` for a genuine follow-up — it keeps Codex's own context and avoids paying
to re-read the repo:
```bash
codex-run -C /path/to/repo -r "Now show me the minimal patch for that."
```
## Rules
1. **Bash timeout ≥ 600000 ms** (10 min) — Codex explores the repo before answering.
2. **Report its answer as Codex's**, not as your own conclusion. Quote it, then add
your own take separately if you disagree — the value here is having two
independent opinions, and blending them destroys that.
3. **Sanity-check it.** Codex can confidently cite files or lines that don't exist.
Verify any specific claim against the repo before acting on it.
4. **Read-only unless told otherwise.** Don't pass `-s workspace-write` on your own
initiative.
5. Each call spends the user's ChatGPT plan quota. One question, one call.
© Sateezg · MIT · Sateezg/codex-bridge