---
name: fallow-rs-perf-loop
description: "Iteratively optimize Fallow performance with stable benchmarks, before-and-after evidence, and correctness gates."
license: "MIT"
metadata:
  decimalai-skill-id: "08ee6647-aa90-4a69-b1ed-e50c1e316508"
  decimalai-version-id: "e828d56d-abe3-45f2-a721-bc62ed80200b"
  decimalai-version: "4"
  decimalai-body-sha256: "07fd5daa46ed7244acaec16697e6948988f145c92fc86e5df5a415dce77969b9"
  decimalai-name: "fallow-rs/perf-loop"
  decimalai-format: "1"
  decimalai-invocation: "model"
  decimalai-category: "coding"
  decimalai-source: "https://app.decimal.ai/s/fallow-rs-perf-loop@4/SKILL.md"
---
<!-- Generated from .agents/skills. Do not edit. -->

# Performance loop

1. Choose a stable benchmark and preserve its identity and workload. Before
   you freeze it, prove that it can see the problem: a known slowdown must
   move the result.
2. Record a statistically useful baseline. Set a target and a minimum number
   of rounds. Do not stop before that minimum unless you reach the target.
3. Profile the hot path before editing.
4. Implement one bounded optimization.
5. Re-run the same benchmark and correctness checks.
6. Keep the change only when the improvement is reproducible and no contract
   regresses.
7. Use a new benchmark identifier for a materially different workload.
8. Run `review`.

Do not report performance gains from debug builds or incomparable fixtures.