CtrlK
BlogDocsLog inGet started
Tessl Logo

getlarge/legreffier

LeGreffier mode: verify identity, sign commits with MoltNet diary, investigate past rationale via signed diary search

90

2.64x
Quality

90%

Does it follow best practices?

Impact

90%

2.64x

Average score across 5 eval scenarios

SecuritybySnyk

Advisory

Suggest reviewing before use

Overview
Quality
Evals
Security
Files

task.mdevals/scenario-2/

Incident Documentation System

Problem/Feature Description

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.

Output Specification

Create the following files:

  1. incident-recorder.ts — A TypeScript module that exports:

    • An IncidentRecord type with all required fields for capturing an incident
    • A createIncidentRecord(input: RawIncident): IncidentRecord function
    • A classifyIncident(description: string): IncidentCategory function
    • A shouldRecordIncident(context: IncidentContext): boolean function that determines whether an incident warrants recording
  2. incident-relations.ts — A TypeScript module that exports:

    • Types for relations between incident records and other records in the system
    • A suggestRelations(incident: IncidentRecord, existingRecords: IncidentRecord[]): SuggestedRelation[] function
  3. entry-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.

evals

tile.json