# Evaluation Contract

Each evaluation must be Markdown with YAML frontmatter containing:

- `artifact_id`
- `evaluator`
- `evaluated_at`
- `scores`

Scores should include groundedness and usefulness on a 0 to 5 scale.