Every section earns its place. The tables are dense with information, the worked example is essential for demonstrating the methodology, and there's no explanation of concepts Claude already knows (no 'what is profiling' preamble).

Provides specific tools for each scenario, executable code in the worked example, concrete commands (py-spy top), and a clear output format template. The lever table gives specific guidance on when each optimization applies.

Clear 4-step sequence (question → profile → pick lever → measure again) with explicit validation checkpoints. Step 3 emphasizes the feedback loop of 'change one thing, measure again' and the output format requires before/after measurements and behavior verification.

Well-organized with clear sections (Step 0-3, worked example, traps, do-nots, output format). Content is appropriately self-contained for a skill of this scope. The reference to behavior-preservation-checker is a clean one-level-deep pointer.

Lists multiple specific concrete actions: 'identifying the actual bottleneck', 'choosing the right optimization lever', and 'measuring the result'. These describe a clear methodology with distinct steps.

Clearly answers both what ('Optimizes code for performance by identifying the actual bottleneck, choosing the right optimization lever, and measuring the result') AND when ('Use when a specific operation is too slow, when a profiler has pointed at a hot path, or when the user asks to make something faster').

Includes natural keywords users would say: 'too slow', 'profiler', 'hot path', 'make something faster'. These cover common variations of how users express performance concerns.

Clear niche focused specifically on performance optimization with distinct triggers like 'profiler', 'hot path', 'too slow', and 'faster'. Unlikely to conflict with general coding or debugging skills.