What problem does it solve? When developing PyPTO operators on Ascend NPU, developers encounter cryptic error codes (F00001, FC4000, etc.) and runtime failures without knowing the root cause or fix. This Skill provides a structured serial lookup over a curated knowledge base of verified fixes and official documentation routes. ## Core Features & Use Cases - Experience Table Lookup: Searches an indexed experience table by error code plus secondary keyword, then drills into 8 classified component files (external, function, machine, pass, matmul, vector, view_op, codegen) to return symptom, root cause, and fix. - Problem Lookup Fallback: When the experience table misses, queries four parallel indexes (keyword index, error-code routing, runtime symptoms, quick-fix table) that route to official troubleshooting documentation chapters and URLs. - Multi-Entry Disambiguation: When one error code maps to multiple entries, returns all candidates for the caller to select based on actual logs instead of guessing. - Use Case: A developer hits F40005 TENSOR_MEMORY_ALLOCATION during PyPTO kernel compilation; the Skill locates the UB/L0B/L1 overflow entry and returns the diagnosis steps and tile-shape fix. ## Quick Start Ask the assistant to look up error code F00001 in the PyPTO experience table and return the matching symptom, root cause, and fix.
