What problem does it solve? AI-assisted work often produces non-trivial decisions whose reasoning, evidence, and uncertainty are lost once the session ends. This Skill preserves those judgments in a standardized, reviewable log so decisions remain inspectable and reusable later. ## Core Features & Use Cases - Structured Judgment Records: Captures the decision, decision status (proposed, accepted, rejected, unknown, deferred), evidence, evidence type, confidence, alternatives, and open questions using a canonical schema. - Evidence Classification: Separates observed facts from inferred conclusions and classifies evidence as deterministic, context_extracted, or inferred, downgrading weakly supported conclusions to unknown. - Use Case: After reviewing a code change and deciding to reject it based on partial evidence, use this Skill to write a judgment log in work/judgments/ that records the decision, the exact evidence paths, remaining open questions, and the next recommended verification step. ## Quick Start Write a judgment log for the decision to reject the mail sender change, using the evidence in the review notes and marking confidence as medium.
