# Pprose Eval

> Score one practical document with metrics and rubric grading; read-only on source. Use when asked to score, evaluate, grade, rubric-check, or measure quality.

- Skill: `jlevy/pprose-eval-2` (Agent Skill)
- Install (CLI): `npx get-skill add jlevy/pprose-eval-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jlevy/pprose-eval-2/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: jlevy (https://skillmd.com/u/jlevy)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/jlevy/pprose-eval-2

---

<!-- DO NOT EDIT: generated by `pprose install` (format=f01) -->

> Run pprose as `pprose <command>` if on PATH, else `uvx pprose@0.1.0 <command>`
> (zero-install via uv).
> Run `pprose --help` for every command, `pprose skill --list` for the other Practical
> Prose skills, and `pprose shortcut --list` / `pprose guidelines --list` /
> `pprose runbook --list` for on-demand playbooks, style guides, and procedures.

# Evaluate One Practical Prose Document

This is read-only on the source document.
It writes eval report files.

Use it when the user wants a formal Practical Prose evaluation of one document.

For the repo’s baseline batch evals, run `pprose runbook practical-prose-baseline-evals`
and default to running all third-party and self-eval artifacts unless the user asks for
a subset.

## Inputs

- Path to one Markdown artifact.
- Artifact label.
- Scope class: `status`, `memo`, `brief`, `deep_research`, or `design_doc`.
- `ANTHROPIC_API_KEY` for model scoring, unless the user asks for dry-run or manual
  scoring only.

## Steps

1. Run `pprose runbook practical-prose-eval-single` for the full procedure.

2. Generate the eval stub:

   ```bash
   pprose report from-metrics path/to/artifact.md --label NAME --scope-class brief --out artifact.eval.md
   ```

3. Inspect deterministic metrics:

   ```bash
   pprose metrics path/to/artifact.md --format yaml
   ```

4. Score the qualitative dimensions (`--model` is required; run
   `pprose score --list-models` for choices, or pass `--dry-run` to skip the API call):

   ```bash
   pprose score artifact.eval.md --model opus
   ```

5. Validate the result:

   ```bash
   pprose report validate artifact.eval.md --complete
   ```

## Output

Return the eval report path, validation result, and any scoring or alignment issues that
need human review.

