ralphex-review
Use Cautionnpx machina-cli add skill salmundani/ralphex/ralphex-review --openclawRalphex Review: Iterative Code Review with Codex
Execute an iterative review loop where Codex reviews your existing code, you (Claude Code) filter the suggestions and implement fixes, then Codex reviews again. Loop until there is no relevant corrections to implement in Codex's review.
Step 1: Read Settings
- Check if
.claude/ralphex.local.mdexists in the project root. - If it exists, read it and parse the YAML frontmatter to extract:
base_branch: The branch to diff against (default:main)codex_model: The Codex model to use (REQUIRED - if missing, ask the user)codex_reasoning_effort: Reasoning effort for Codex reviews (default:high)
- If the file does not exist, first run
command -v codexvia Bash. If the command exits with a non-zero status (codex not found), inform the user: "Codex CLI is not installed. Install it from https://github.com/openai/codex and try again." and stop. Otherwise, ask the user:- What base branch to diff against (suggest
main) - What Codex model to use (
gpt-5.3-codex,gpt-5.2-codex,gpt-5.3-codex-spark) - What reasoning effort to use for reviews (suggest
high; valid values:low,medium,high,xhigh) Then create.claude/ralphex.local.mdwith their answers.\
- What base branch to diff against (suggest
- Check for uncommitted changes by running
git status. If there are uncommitted changes, useAskUserQuestionto present these options: a. Commit modified tracked files: Rungit add -uto stage all modified tracked files. Ask the user for a commit message, then commit with that message. b. Commit all changes (modified + untracked): Rungit add -Ato stage everything including untracked files. Ask the user for a commit message, then commit with that message. c. Stash changes: Generate a unique stash name by runningdate +%sand using"ralphex-review-stash-{timestamp}". Rungit stash push -m "ralphex-review-stash-{timestamp}". Store the full stash message for later cleanup. d. Cancel: Stop the review. - Resolve the review target: Run
git rev-parse --abbrev-ref HEADto get the current branch. Compute areview_targetvalue:- If the current branch is different from
base_branch, setreview_target=base_branch. - If the current branch equals
base_branch, check iforigin/{base_branch}exists by runninggit rev-parse --verify origin/{base_branch}. If it exists, setreview_target=origin/{base_branch}. If it does not exist, inform the user that there is no remote to compare against and stop. - Verify there is actually a diff by running
git diff --stat {review_target}...HEAD. If the diff is empty, inform the user there are no changes to review and stop.
- If the current branch is different from
Store review_target, codex_model, and codex_reasoning_effort for use throughout the workflow. Use review_target (not base_branch) in all subsequent git and Codex commands.
Step 2: Codex Code Review
Run Codex to review the branch changes:
-
Run this command via Bash, replacing
{review_target},{codex_model}, and{codex_reasoning_effort}with the resolved values from Step 1, unless the user prompt "$ARGUMENTS" says otherwise:codex exec --sandbox read-only -m {codex_model} -c model_reasoning_effort="{codex_reasoning_effort}" -o .claude/ralphex-review.txt "Review the committed changes of this branch against {review_target}. $ARGUMENTS" 2>/dev/null -
Check the exit code. If the command exits with a non-zero status, inform the user that Codex failed (include the exit code). Perform Cleanup: Restore Stashed Changes if a stash was created in Step 1, then stop.
Step 3: Evaluate Review
Read the contents of .claude/ralphex-review.txt.
If corrections exist, critically evaluate each one
Do NOT blindly accept all suggestions. Codex can be wrong, pedantic, or suggest changes outside the task scope. For each suggestion in the review, independently evaluate and classify it:
- Accept: Real bugs, security issues, correctness problems, or clear improvements directly related to the task. These will be addressed in the next iteration.
- Reject: Factually wrong suggestions, misunderstandings of the code or requirements, or purely stylistic preferences that don't improve correctness.
- Defer: Valid observations that are out of scope for the current task (e.g., pre-existing issues, unrelated refactoring suggestions).
Display your verdicts to the user in this format:
**Iteration {N} - Codex Review Evaluation:**
- ✅ Accept: {summary of accepted suggestion}
- ❌ Reject: {summary and reason for rejection}
- ⏭️ Defer: {summary and reason for deferral}
Then decide:
- If all suggestions are rejected or deferred → treat this as a clean review. Inform the user. Perform Cleanup: Restore Stashed Changes if a stash was created in Step 1, then stop.
- If any suggestions are accepted → delete
.claude/ralphex-review.txtusing Bash:rm .claude/ralphex-review.txt, then proceed to Step 4: Implement with only the accepted corrections. Do not address rejected or deferred items.
If clean
- Inform the user that Codex approved the implementation
- Display the total number of iterations it took
- Delete
.claude/ralphex-review.txtusing Bash:rm .claude/ralphex-review.txt - Perform Cleanup: Restore Stashed Changes if a stash was created in Step 1.
Step 4: Implement
Implement the accepted corrections:
- For each accepted correction, explore the relevant code to understand the context.
- Make the necessary code changes. Keep changes focused: only address accepted corrections, do not refactor unrelated code.
- Verify changes are consistent with the rest of the codebase.
Step 5: Commit
Create a clean git state for Codex to review again:
- Stage all changed files with
git add(be specific about which files. Do NOT usegit add .orgit add -A). - Commit with a descriptive message summarizing what was fixed based on the Codex review.
- Verify changes are committed with
git status.
After committing, go back to Step 2: Codex Code Review for the next iteration.
Cleanup: Restore Stashed Changes
If a stash was created in Step 1, restore it before stopping:
- Run
git stash listand find the entry whose message contains the stash name stored in Step 1 (e.g.,"ralphex-review-stash-{timestamp}"). - Extract its
stash@{N}reference from the matching line. - Run
git stash pop stash@{N}. - If the pop fails due to conflicts, inform the user that the stash could not be applied cleanly and tell them to resolve conflicts and run
git stash popmanually.
Important Rules
- Never skip the commit step. Codex reviews committed diffs.
- Track iteration count. Display it with each review cycle so the user knows progress.
- The loop is fully automatic. No user approval is needed between iterations. Only stop when Codex gives a clean review, or all suggestions are rejected/deferred.
- If Codex CLI fails (command not found, network error, etc.), inform the user and stop. Do not retry automatically.
- The workspace must be clean before the first review. Codex reviews committed diffs. If there are uncommitted changes, offer to commit them, stash them, or cancel.
- If changes were stashed, always restore them. Every exit path must perform the Cleanup: Restore Stashed Changes step before stopping.
- Clean up
.claude/ralphex-review.txtafter reading it in every iteration, whether the review is clean or not.
Source
git clone https://github.com/salmundani/ralphex/blob/main/ralphex/skills/ralphex-review/SKILL.mdView on GitHub Overview
Ralphex Review automates an iterative code-review loop where Codex analyzes your changes, Claude Code filters and implements fixes, then Codex reviews again until no relevant corrections remain. It relies on a local config at .claude/ralphex.local.md to set base_branch, codex_model, and codex_reasoning_effort, and uses git diff targets to drive the review.
How This Skill Works
The workflow reads settings from .claude/ralphex.local.md (or prompts to create it), computes a review_target based on your current branch and remote, and runs Codex to review the committed changes. If corrections exist, Claude Code applies the fixes and Codex reviews again, looping until there are no further relevant corrections to implement.
When to Use It
- You want to review a feature or patch branch and surface actionable fixes from Codex.
- You plan to filter Codex suggestions with Claude Code before applying changes.
- You need to ensure there is a meaningful diff against base or origin before starting the review.
- Your project uses a .claude/ralphex.local.md configuration and the Codex CLI is available.
- You want to iterate until the review reports no further corrections to implement.
Quick Start
- Step 1: Create or verify .claude/ralphex.local.md with base_branch, codex_model, and codex_reasoning_effort. Ensure Codex CLI is available.
- Step 2: Run Codex review on the current branch against the computed review_target, saving results to .claude/ralphex-review.txt.
- Step 3: If .claude/ralphex-review.txt lists corrections, apply the fixes and re-run the review until no further corrections remain.
Best Practices
- Create and maintain .claude/ralphex.local.md with base_branch, codex_model (required), and codex_reasoning_effort (default high).
- Work in a clean git state or use the provided options to handle uncommitted changes before starting.
- Ensure review_target is correctly determined (base against the current branch or origin) and that a diff exists.
- Apply Codex suggestions via Claude Code thoughtfully, then re-run the review to validate changes.
- Document final fixes with clear commit messages and summarize changes after the loop completes.
Example Use Cases
- Refactoring a Python data processing module and iteratively polishing edge-case handling.
- Cleaning up a JavaScript utility library with multiple small fixes surfaced by Codex.
- Auditing a Bash automation script for robustness and safer error handling.
- Reviewing SQL migration changes to ensure correctness and idempotence.
- Tuning a CI workflow file to improve reliability and reduce flaky runs.