Adversarial hypothesis reviewer — systematically attacks theories and root cause analyses to find weaknesses before they find you.
Install via Claude Code's plugin system:
/plugin marketplace add openshift-eng/edge-tooling
/plugin install challenge
During any conversation where you've formed a hypothesis or root cause theory:
/challenge
Claude will:
- Extract the hypothesis from conversation context
- Break it into falsifiable claims
- Actively search the codebase for counter-evidence
- Present structured counter-arguments (CA-1, CA-2, ...) ranked by severity
- Suggest experiments to confirm or deny each challenge
- After forming a root cause theory during debugging
- After proposing an explanation for a CI failure or test regression
- After drawing conclusions from log analysis
- Any time you want your reasoning stress-tested before acting on it
A structured adversarial review containing:
- Counter-arguments with severity/likelihood ratings
- Evidence gaps — claims that couldn't be verified
- Assumptions inventory — implicit assumptions that may not hold
- Alternative explanations with distinguishing tests
- Next steps ordered by information value
- Investigation log of all searches performed
- Claude Code: >= 1.0.0
- Category: debug
fonta-rh