gate-keeper-bench

What problem does it solve?

This skill automates end-to-end benchmarking of Gate Keeper detection across curated fixtures to quantify accuracy, false positives, and stability.

Core Features & Use Cases

• Automates end-to-end benchmarking of Gate Keeper across spec-review and code-review fixtures, collecting per-run verdicts and summary metrics.
• Handles fixture loading, mode filtering, manifest validation, and deterministic 3-run evaluation per fixture with robust error handling and progress reporting.
• Produces a comprehensive report with per-fixture results, per-mode metrics, and category-detection insights for known-bad fixtures.

Quick Start

Run the benchmark to execute three evaluations per fixture and generate the final report.