Challenger

Multi-agent adversarial plugin that stress-tests code decisions, architecture plans, and reasoning through evidence-backed challenge rounds. Four specialized agents (Skeptic, Sentinel, Architect, Pragmatist) take turns poking holes in your conclusions using actual codebase evidence — reading code, searching patterns, and checking git history. Auto-scales intensity from quick sanity checks to brutal multi-round reviews based on stakes. Includes a passive advisory agent that suggests challenging d