What problem does it solve? After an AI research route finishes answering a question over a document collection, there is no built-in way to judge whether the process itself was sound or whether the framework should improve for future requests. This Skill closes that loop by auditing completed routes and recording process-quality findings. ## Core Features & Use Cases - Route Process Audit: Reads the original prompt, goal artifact, chain, produced artifacts, and verifier outcome, then classifies findings under categories like integrity_issue, route_selection_issue, evidence_handling_issue, and efficiency_issue. - Edit Decisioning: Decides between no_edit and edit_recommended, naming the exact control or doc files and the smallest safe change when an edit is justified. - Provenance Footer & Cleanup: Appends a Search Provenance footer built from the session evidence packet to the verified report, archives evidence packets to .spinosa/archive/, and moves session-scoped intermediate files to .trash/. - Use Case: After a verified answer report is produced for a research question over interview transcripts, run this Skill to write an audit report to agent_reports/e_{session_id}.md and decide whether the framework's routing or evidence-handling rules need a targeted fix. ## Quick Start Audit the completed route for this session and write the evaluation report with an edit decision to agent_reports.
