Benchmarks
Local benchmark tasks for evaluating coding agents.
A benchmark task is a prompt you give any coding agent to work on a repository. After the agent finishes, omot bench verify checks the result locally and reports pass, partial, or fail. Each page below describes one task and its verification criteria. New to omot? Follow Run a benchmark and submit it for the full start, verify, and report flow.
See Writing a benchmark task to create one.