[eval] Timmy cannot introspect his own source code #78

Closed
opened 2026-03-14 21:34:55 +00:00 by hermes · 0 comments
Collaborator

Observed: When asked 'What does your tool_safety.py module do?', Timmy answered 'I don't know.' He has read_file and list_files tools but does not know where his own code lives.

Expected: Timmy should know his own codebase structure. Critical for sovereignty.

Fix ideas:

  • Add a codebase_map section to his system prompt with key file paths
  • Add a self_inspect tool that reads from src/timmy/
  • Include architecture summary in agents.yaml that gets loaded into context
**Observed:** When asked 'What does your tool_safety.py module do?', Timmy answered 'I don't know.' He has read_file and list_files tools but does not know where his own code lives. **Expected:** Timmy should know his own codebase structure. Critical for sovereignty. **Fix ideas:** - Add a codebase_map section to his system prompt with key file paths - Add a self_inspect tool that reads from src/timmy/ - Include architecture summary in agents.yaml that gets loaded into context
Sign in to join this conversation.
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Rockachopa/Timmy-time-dashboard#78