Five questions → are you running a multi-vendor agentic fleet, and if so, which cost shape is hitting you hardest? Maps to the Multi-Vendor Fan-Out Field Guide.
1. How many distinct LLM vendors are you currently running workers against in production?
2. Do you operate a supervisor process you wrote that dispatches tasks to worker processes?
3. Have you observed workers retrying the same task with slightly different parameters, where the supervisor failed to catch the collision?
4. Where does your monthly API spend across all vendors sit?
5. When workers run in parallel, do you have any instrumentation that measures the ratio of (worker retries caught by supervisor) vs (worker retries that completed before being caught)?
This diagnostic is based on the Multi-Vendor Fan-Out Field Guide and the architectural discussion in anthropics/claude-code#64080 with @palios-taey. It does not collect or transmit your answers — all scoring runs in your browser. The recommendations assume you have read the field guide; they are pointers, not a substitute for the guide's full articulation of the cost shapes and trade-offs.