skill-reporter

Generate standalone HTML reports for skill iteration workspaces.

What it does

skill-reporter reads an iteration directory containing benchmark metadata, grading results, timings, and output code, then emits a single self-contained HTML report such as iteration-3.html.

The generated report includes:

• benchmark summary cards
• analyst notes
• per-eval overview table
• assertion-by-assertion grading comparison
• side-by-side Shiki-rendered code panes

Intended package usage

After the package is published to npm, the intended interface is:

npx skill-reporter /path/to/iteration-3

By default this writes:

/path/to/iteration-3.html

You can also provide a custom output path:

npx skill-reporter /path/to/iteration-3 --out /tmp/iteration-3.html

Local development

bun install
bun run dev ./path/to/iteration-3
bun run build
node dist/cli.js ./path/to/iteration-3

Expected input layout

iteration-N/
  benchmark.json
  eval-name/
    eval_metadata.json
    with_skill/
      grading.json
      timing.json
      outputs/*
    without_skill/
      grading.json
      timing.json
      outputs/*