The Behaviour Debugging Prompt

My agent is behaving badly. Help me find the cause.

The behaviour: [DESCRIBE PRECISELY — what it does, how often]
System prompt: [PASTE]
Tools: [LIST]
Example transcript: [PASTE A RUN WHERE IT WENT WRONG]

Work through:
1. At which step does the behaviour first appear?
2. What information did the agent have at that point?
3. What did it appear to conclude, based on its actions?
4. Is this a prompt problem, a tool problem, a schema problem, a
   context problem, or a model capability problem? Argue for one.
5. What specific change would fix it?
6. What would that change break?
7. How would I verify the fix without shipping it?

Be specific about the diagnosis. "Improve the prompt" is not a
diagnosis. Name the instruction that is missing or wrong.