Main entry point.
md-evals [OPTIONS] COMMAND [ARGS]Scaffold eval.yaml and SKILL.md.
md-evals init [DIRECTORY] [OPTIONS]| Option | Description |
|---|---|
--force, -f |
Overwrite existing files |
Run evaluations.
md-evals run [OPTIONS]| Option | Alias | Description |
|---|---|---|
--config, -c |
Config file | Default: eval.yaml |
--treatment, -t |
Treatment(s) | Comma-separated or wildcard |
--model, -m |
Override model | Model name |
--count |
Repetitions | Number of runs |
-n |
Parallel workers | Number of parallel workers |
--output, -o |
Output format | table, json, or markdown |
--no-lint |
Skip linting | Skip skill validation |
--verbose, -v |
Verbose output | Detailed results |
Validate SKILL.md.
md-evals lint [SKILL_PATH] [OPTIONS]| Option | Alias | Description |
|---|---|---|
--fail, -f |
Exit with error | Default: true |
--verbose, -v |
Show details |
List treatments and tasks.
md-evals list [OPTIONS]| Option | Alias | Description |
|---|---|---|
--config, -c |
Config file | Default: eval.yaml |
--treatments, -t |
List treatments | |
--tasks |
List tasks |
Show version.
md-evals version