second-opinion

You can run an LLM code review on uncommitted changes or branch diffs by invoking external tools like Codex or Gemini. The skill processes the diff scope you select and presents findings focused on security, performance, and error handling.

Yes, you can use both Gemini and Codex to review uncommitted changes. The skill supports running both external LLM CLIs in parallel and summarizes the results, highlighting where the two models agree or differ on your code.

LLM code review can check for vulnerable dependencies by applying diff-aware dependency scanning. This optional security scanning runs specifically when your changes touch dependency manifests, ensuring targeted vulnerability detection.

To align external AI code review results with your project conventions, the skill incorporates standards from CLAUDE.md or AGENTS.md files. When present in your repository, these files guide the LLM to match your established coding standards.

External LLM code review handles empty or large diffs using built-in guardrails. These safety checks prevent the tool from processing empty changes and manage large diffs to ensure the external model review remains effective and within limits.

Yes, you need Codex or Gemini installed to get a second opinion on your code. The skill invokes these external models via their respective CLIs, requiring you to select a review tool and diff scope to execute the independent code review.