feat: add codex-review plugin for automated code review v1.0.0#15
Conversation
Summary of ChangesHello, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request integrates a new Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here. Footnotes
|
There was a problem hiding this comment.
Code Review
This pull request introduces a new codex-review plugin for automated code reviews. The plugin is well-structured with a command, a skill, and clear documentation. The implementation thoughtfully includes an iterative fix-and-review loop with safety guards. My review has identified a significant dependency risk due to the use of a deprecated tool, and a minor inefficiency in the review loop's logic. Addressing these points will improve the plugin's long-term viability and performance.
First command-based plugin in the marketplace. Provides /codex-review slash command with auto-detection of review mode (--uncommitted, --base, --commit) and iterative fix-and-review loop with three anti-loop safety guards. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
… cycles - Severity informs priority but doesn't mechanically determine the action - Redefine "no progress" as all findings dismissed or already fixed - Increase max loop cycles from 3 to 4 Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Remove content that overlaps with the command (background execution, silent fallback, mode detection, when-to-review table). Skill now only covers the review mandate, own-code rule, and severity guidance. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
…ure v2.0.0 Separate concerns into command (execution), skill (autonomous behavior), and agent (subagent delegation) — matching the proven coderabbit plugin pattern. The v1 monolithic command was brittle and hard to trigger proactively. - commands/review.md: clean execution wrapper with context injection - skills/code-review/SKILL.md: when/how to review, autonomous fix loop - agents/code-reviewer.md: specialized subagent for thorough analysis - Removes old codex-review.md command and review-workflow skill Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
59e0535 to
31ad4b3
Compare
…command - `committed` mode now uses `codex review` (no flags) instead of falling back to --uncommitted when no PR exists — fixes wrong-diff bug - `--base` passes the ref as-is instead of prepending origin/, supporting multi-remote setups Found by codex review during plugin testing. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Both plugins installed = both would try to review autonomously after every code change. Match coderabbit's reactive pattern instead — only review when the user explicitly asks. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Summary
codex-reviewplugin v2.0.0 — restructured to match the proven coderabbit plugin architectureWhat changed from v1
codex-review.md(monolithic: review + fix loop + anti-loop)review.md(clean execution wrapper with context injection)review-workflow/SKILL.md(thin: just policy/severity)code-review/SKILL.md(full review workflow with fix loop)code-reviewer.md(specialized subagent for thorough analysis)Plugin structure
Design decisions
--basepasses refs as-is — no hardcodedorigin/prefix, supports multi-remote setupsTest plan
prek run --all-filespasses/codex-review:reviewcommand works (tested: auto-detected PR, ran review, found real bugs)🤖 Generated with Claude Code