From claude-connoisseur
Sends plans, code, diffs, or approaches from context to OpenAI Codex CLI for independent review. Useful for validating implementations, architecture, or security before committing changes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-connoisseur:codex-reviewThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
When invoked, send the relevant context — plans, approaches, code, or any combination — to `codex` CLI for an independent review.
When invoked, send the relevant context — plans, approaches, code, or any combination — to codex CLI for an independent review.
Files and instructions: $ARGUMENTS
This skill covers both code and plans/approaches:
Adapt what you send based on what the user asks Codex to review.
The arguments are free-form and flexible. They may contain:
@auth.ts review this file for security issuesParse the arguments to determine what context to gather and send to Codex.
/codex-review review my current plan — review the plan from the current conversation/codex-review review the code I just wrote — review recent code changes/codex-review review the approach and the implementation together — review both plan and code/codex-review @auth.ts review this file for security issues — review a specific file with focus/codex-review review the last commit — review the most recent commit diffGather context: Collect the relevant material to send for review based on what the user asked. Match the scope of what you send to the scope of the ask — if the user points at a single function, send that function and its immediate dependencies, not the entire file history or conversation. A quick opinion needs adequate context; a full architecture review needs more.
git diff) and a brief description of intent — not every file in the repoBuild the prompt: Construct a clear prompt for Codex that includes:
Run codex: Write the full prompt to a temporary file, then pipe it to the helper script:
echo "Review the following <code/plan/approach> for a software engineering task and provide feedback.
## Project Context
<project context>
## What to Review
<code, plan, approach, or combination>
## Please provide:
1. Overall assessment
2. Potential issues or risks
3. Suggested improvements
4. Anything missing or overlooked
## Output rules
- Be concise
- Do not echo back the entire code or plan you were given
- Short code snippets are fine to illustrate a point, but never include entire files
- If you explore the codebase, summarise what you found — do not paste what you read" | skills/codex-review/run-codex.sh
Important:
echo "..." | skills/codex-review/run-codex.sh — never use heredocs (indentation causes hangs)codex is installed and exits with a clear error if notReport back: Present the Codex feedback to the user clearly:
npx claudepluginhub eugeniosegala/claude-connoisseur --plugin claude-connoisseurCross-model review using OpenAI Codex to independently verify plans or code diffs, iterating up to 5 rounds. Useful for architecture decisions, non-trivial refactors, and critical config changes.
Invokes Codex to review implementation plans for gaps, risks, suboptimal steps, and alternatives. Triggered after Claude plans or on requests for second opinions.
Runs official Codex code review on code changes, diffs, branches, or commits, then double-checks each finding by reading source code and classifying agree/disagree/nuance.