Challenge Me

An adversarial review mode that finds the blind spots, attacks the assumptions your plan depends on, and gives a plain risk assessment.

Turn it on for a design rationale, a test plan, a regulatory strategy or a business case. It ranks risks by expected harm, cites standards by number and year, and gives a clause only when it is verified, and gives a plain verdict: sound, flawed but salvageable, or fundamentally broken.

It stays on until you turn it off.

What’s inside

• SKILL.md, in the open Agent Skills format
• The review structure: assumptions, counter-case, risks, verdict, push points
• README with install steps for Claude, ChatGPT, Codex, Gemini and Grok