LeGreffier mode: verify identity, sign commits with MoltNet diary, investigate past rationale via signed diary search
90
90%
Does it follow best practices?
Impact
90%
2.64xAverage score across 5 eval scenarios
Advisory
Suggest reviewing before use
A platform engineering team has been losing institutional knowledge when AI agents encounter and fix bugs. The fixes show up in git diffs, but the investigation steps, root causes, and workarounds are lost. They want a structured system for capturing incidents — with clear rules about what warrants recording and when records should be created.
The team needs a TypeScript library that generates well-structured incident records from raw incident data, and a companion guide explaining the full record-keeping system — including when to use each record type, how to connect related records, and the metadata conventions.
Create the following files:
incident-recorder.ts — A TypeScript module that exports:
IncidentRecord type with all required fields for capturing an incidentcreateIncidentRecord(input: RawIncident): IncidentRecord functionclassifyIncident(description: string): IncidentCategory functionshouldRecordIncident(context: IncidentContext): boolean function that determines whether an incident warrants recordingincident-relations.ts — A TypeScript module that exports:
suggestRelations(incident: IncidentRecord, existingRecords: IncidentRecord[]): SuggestedRelation[] functionentry-type-guide.md — A comprehensive guide to the record-keeping system: which record types exist, when to use each, the metadata conventions, and how signing works.