What problem does it solve? When an agent authors a platform-ecosystem prior-art survey (vocabulary maps, search outputs, extract records, decision indexes), it cannot reliably grade its own work. This Skill is the judgment half of a two-part gate: it reviews such artifacts against a fixed set of numbered conditions and decides whether they can be built on, without repeating what the deterministic validator already checks. ## Core Features & Use Cases - Condition-based review: Judges artifacts against 31 numbered conditions (C1–C31) covering slug provenance, applicability verdicts, verbatim query recording, zero-hit cells, enumeration framing, reversibility calls, authority bands, and divergence handling, all defined in references/conditions.md. - Evidence-grounded findings: Requires every finding to be grounded in the artifact, the producer's schemas, the source registry, or the angle reference; anything ungrounded is demoted to an Observation. - Proportionality and upstream labeling: Refuses to revise thin-but-honest results, and labels findings whose remedy lies outside the artifact as UPSTREAM: with the exact file and field named. - Use Case: After running the platform-ecosystem-prior-art-survey producer skill, hand its search output and the wave-0 vocabulary map to this reviewer; it returns findings naming their condition and a single terminal VERDICT: approve or VERDICT: revise line. ## Quick Start Review this platform-ecosystem prior-art search output against the conditions and give me the verdict.
