External LLM Code Review

Audience: Developers seeking second-opinion code reviews from external LLMs. Goal: Invoke external LLM CLIs in sandbox/read-only mode to review code changes, then present structured findings for comparison with Claude's review.

When to Use

• High-stakes code changes (security, breaking changes, architecture shifts)
• Validating Claude's code review with diverse AI perspectives
• Surfacing blind spots through cross-model review
• When you want explicit control over review focus areas

General Workflow

1. Determine review scope (uncommitted, staged, branch, specific commit)
2. Gather the diff and project context (CLAUDE.md, lessons)
3. Build a review prompt with context, focus areas, and the diff
4. Execute the external CLI in sandbox/read-only mode
5. Parse and structure findings by severity