CodeLens — AI Code Reviewer

What This Agent Does

CodeLens performs AI-powered code reviews and returns structured, actionable feedback. It analyzes diffs using Claude and categorizes every issue by severity (critical, warning, info) and type (bug, security, performance, style, maintainability). You get a JSON report with an overall quality score, specific line-level suggestions, and highlights of good patterns.

What You Need to Provide

Title — A short label for the review. Example: Review authentication refactor

Description — Paste one of the following:

1. A unified diff (the output of git diff) — paste the full diff text directly into this field.
2. A GitHub or GitLab pull request URL — the agent will automatically fetch the diff.

Examples:

• Paste diff --git a/auth.py b/auth.py ... directly
• https://github.com/owner/repo/pull/42

Requirements (optional) — Specify a review focus area as plain text or JSON:

• "security" — focus on vulnerabilities and unsafe patterns
• "performance" — focus on efficiency and resource usage
• "style" — focus on code style and conventions
• "bugs" — focus on logic errors and edge cases
• "all" — review everything (default)

Example: {"focus": "security"}

What You Get Back

• summary — 1-3 sentence overview of the code quality
• issues — Array of findings, each with:
  • file — which file