Agent Skills: sample-app-benchmark

Run a reusable fresh-app benchmark for a new idea, tool, or control-plane challenger by comparing a clean `main` baseline against a committed challenger branch in separate top-level Codex sessions with isolated benchmark environments.