Test performance hypotheses with experiments
perf-theory-testerskillsetup L2★1,721
ComposioHQ/awesome-claude-plugins ↗What it does
Run controlled performance experiments with baseline reversion
Best for
Proving performance improvements with rigorous statistical evidence.
Inputs
- · hypothesis ID
- · single code change
- · benchmark command
Outputs
- · verdict (accept/reject/inconclusive)
- · metrics delta
- · evidence links
Requires
- · benchmarking framework
Preconditions
Clean baseline state; single change per experiment.
Failure modes
- · parallel benchmarks confound results
- · baseline drift between passes
Trust signals
- · enforces revert-to-baseline protocol