skeptic

AI output validation uses cognitive frameworks like Cynefin, Inversion, and Chesterton's Fence to review large plans and diffs. This framework-based analysis identifies gaps, hallucinations, and overreach by checking evidence coverage, surfacing blind spots, and producing structured findings with severity levels and recommended mitigations.

Reviewing AI diffs and tool outputs involves applying evidence-focused evaluation to highlight missing data and untested edge cases. The process yields structured findings with severity levels and concrete actions, linking identified gaps and overreach directly to risk and governance workflows for mitigation.

The best way to validate AI outputs is applying framework-driven critique using Cynefin, Inversion, Chesterton's Fence, and Devil's Advocate. This skeptical review emphasizes evidence and coverage, surfacing blind spots in AI-generated plans and producing actionable findings with severity levels to ensure safe conclusions.

This AI validation skill interoperates with other safety and validation skills, such as guard and ship, for end-to-end risk management. It provides framework-based analysis and structured findings with severity levels that link directly into downstream risk and governance workflows.

You need framework-based analysis when AI-generated plans, diffs, and analyses risk gaps, overreach, or hallucinations. Applying evidence-focused evaluation highlights missing data and untested edge cases, providing a skeptical review that emphasizes evidence, coverage, and safe conclusions with actionable mitigations.