capability-diagnostic

To diagnose agent failures, you can use a structured decision tree to identify missing tools, insufficient context, inadequate model capability, or infrastructure issues. This approach isolates the specific root cause behind recurring task failures and schema mismatches.

Repeated enforcement-hook blocks and runtime overages often stem from ambiguous prompt specifications or infrastructure reliability problems. Identifying whether the issue is missing context or inadequate model capability resolves these recurring blocks effectively.

Fixing schema-mismatched outputs requires checking for inadequate model capability or ambiguous specifications. By applying a failure analysis decision tree, you can determine if updating task prompts or escalating model tiers is necessary.

You should escalate model tiers when a failure analysis decision tree indicates inadequate model capability. If updating agent specs or improving task prompts does not resolve the issue, escalating model tiers addresses the capability gap.

Diagnosing agent capability gaps requires no specific dependencies. It relies on a standardized five-step decision tree to analyze failures and output structured Diagnosis, Root cause, and Fix fields for fast handoffs.