Pr Comments

Automates the grind of triaging PR review feedback by fetching all unresolved threads via GraphQL, deduplicating bot spam across CodeRabbit, Gemini, and other CI bots that often flag the same issue, then sorting by severity and confidence. It writes a human-readable fix plan that separates real problems from noise (outdated threads, style nitpicks on config files, bots contradicting your project conventions), waits for your approval, then resolves the junk and spawns parallel subagents to fix the actual issues. The deduplication logic is surprisingly thorough: groups by file and line range, parses severity from each bot's quirky format (emoji badges, SVG alt text, shield images), and takes the highest severity when multiple bots agree. Saves you from manually cross-referencing twenty bot comments to find the three that matter.