Automated Code Review Agent

About

The Automated Code Review skill transforms Claude into a sophisticated quality assurance partner capable of auditing uncommitted changes or specific issue-linked diffs. By leveraging parallel specialized agents—including security scanners and silent failure detectors—the skill provides a comprehensive analysis of code integrity. It minimizes noise by filtering findings through a confidence scoring system, automatically repairing high-certainty bugs while ensuring strict compliance with local project standards defined in CLAUDE.md files.

Key Features

• Deep-dive thorough mode for critical security and silent failure detection
• Confidence-based filtering (80%+) to eliminate false positive warnings
• 1 GitHub stars
• Automated bug fixing for high-confidence (95%+) code violations
• Multi-agent parallel processing for general, security, and logic reviews
• Project-specific compliance checks using root and directory-level CLAUDE.md files

Use Cases

• Conducting deep security audits on sensitive code changes using the --thorough flag
• Enforcing consistent project style and documentation standards across a distributed team
• Performing pre-commit sanity checks to catch logic bugs before they reach CI