Agents & automationTesting & evalsagent

Find why an agent failed

Diagnoses an agent run without turning guesses into fixes.

Editorial illustration for this prompt; the prompt itself does not generate this image.

Ready to use

Prompt

Investigate this failed agent run: [GOAL, INSTRUCTIONS, INPUT, TOOL TRACE, ACTUAL RESULT]. Reconstruct what happened in order. Separate prompt defect, missing tool access, stale data, tool error, unclear authority and evaluation failure. Identify the first observable divergence. Suggest the smallest change and one regression test. Do not modify the live workflow until the cause is supported by evidence.