Skip to main content

Eval Viewer

The action uploads an interactive HTML viewer as a GitHub Actions artifact.

Tabs​

Grading​

Shows per-case results with:

  • Pass/fail badges for each criterion
  • Evidence quotes from the grader
  • Full response output
  • Duration and token count

Benchmark​

Summary table with:

  • Overall pass rate
  • Per-case pass rate with visual bars
  • Time and token usage
  • Total cost

Downloading​

Artifacts appear at the bottom of each workflow run. The filename includes a timestamp and skill name (e.g., 20260317T100000-tf-guide.html).

Disabling​

Set upload-viewer: 'false' in the action inputs to skip artifact upload.