# Benchmarks

This workspace contains deterministic benchmark suites used to evaluate governed agent execution and reproducibility.

Use `npx martin-loop bench --suite under-3-challenge` for the primary public benchmark lane.