eval-runner - React-Craft Evaluation Test Runner
Executes build pipeline against fixtures, applies graders, and produces benchmark reports.
Tags
Updated: 2026-03-24Typical Inputs
Typical Outputs
What this skill does
- Read fixture metadata
- Read fixture input
- Sanitize component names
- Sanitize variant names
- Escape HTML entities
- Validate file paths
- Run build pipeline
- Apply deterministic graders
- Apply LLM judges
- Calculate weighted scores
- Write results file
- Write grader details
Inputs
- Fixture path
- Output path
- metadata.yaml
- input.yaml
- EXPECTED_FINDINGS.md
- GRADERS flag
Outputs
- results.yaml
- grader-details.md
- Pipeline output directory
- Benchmark scores
- Agent metrics
Requirements
- Node.js environment
- TypeScript compiler
- ESLint
- Prettier
- Pipeline agent access
