Adversarial code review for GitHub PRs with: - Cynical internal review phase, professional external output - Severity-based findings with [likely] tag for high-confidence issues - LLM disclaimer in output footer - Safety features: sandboxed execution, large PR warnings, preview before post |
||
|---|---|---|
| .. | ||
| README.md | ||
| review-prompt.md | ||
README.md
Raven's Verdict - Deep PR Review Tool
Adversarial code review for GitHub PRs. Works with any LLM agent.
Prerequisites
ghCLI installed and authenticated (gh auth status)- Any LLM agent capable of running bash commands
Usage
# Claude Code
claude "Review PR #123 using tools/maintainer/pr-review/review-prompt.md"
# Other agents: copy review-prompt.md contents to your agent
See review-prompt.md for full details on output format, safety features, and how it works.