Cost Booster Edit

This is the execution half of the cost-booster pair. Route classifies whether a transform is simple enough for deterministic WASM handling, then this skill actually runs it via agent-booster's apply() function. The benchmark numbers are legitimate: 12/12 correctness on structural transforms like var-to-const or add-error-handling, sub-millisecond median latency, and genuinely zero cost since there's no LLM call. The 1838x speedup claim versus Sonnet is real but only meaningful for bulk edits where you'd otherwise burn tokens on repetitive transforms. Use it when you have many files needing the same mechanical change. Skip it if the edit requires any reasoning about intent or naming.