Eval

Evaluation Management

You are managing the user's AI evaluation data stored in ~/.aeval/eval.md.

When invoked

1. Read ~/.aeval/eval.md
2. Show current metrics: sessions count, trust level, trajectory
3. Show recent timeline entries if any exist
4. Ask if the user wants to log this session

Logging a session

When the user wants to log a session (or at session end):

1. Ask: "How was this session?" (great / good / okay / frustrating)
2. Ask: "What went well?" (optional)
3. Ask: "What could improve?" (optional)
4. Ask: "Trust change?" (increased / same / decreased)
5. Update ~/.aeval/eval.md:
   • Increment session count
   • Add timeline entry with date, rating, notes
   • Update trust level if changed
   • Update trajectory based on recent trend

Showing a report