# Evaluating prose

Model output is a candidate, not a guaranteed improvement. Read the complete result before replacing or publishing anything. A no-op edit can be the correct outcome when the input is already strong.

## Work covenant

Write a short, high-level account of the work before comparing versions. Describe its essential subject, governing idea, emotional or rhetorical destination, human truth, and honesty. Do not turn particular lines, incidents, paragraph shapes, reveal order, devices, target length, or reference-style resemblance into requirements.

Answer `YES` or `NO` for each candidate:

1. **Identity** — Is this recognizably the same work in its essential subject, spirit, and point of view?
2. **Meaning** — Does it preserve the governing idea rather than substitute a different work?
3. **Emotional destination** — Is it still reaching toward the same emotional or rhetorical destination?
4. **Human truth** — Do people, motives, choices, relationships, and consequences remain credible on the work's own terms?
5. **Honesty** — Is it free of inventions that materially misrepresent what the work claims to be true?

Any `NO` means the candidate is a different work for selection purposes, even if some writing is excellent.

## Reading quality

Answer `YES` or `NO` independently:

1. **Compulsion** — Does the prose make you want to continue?
2. **Living voice** — Does it sound like an intelligence speaking rather than generic competent prose?
3. **Emotional effect** — Does it evoke its intended feeling rather than announce it?
4. **Movement** — Does it develop with momentum, discovery, deepening, or turn?
5. **Rhythm** — Do sentence and paragraph rhythms suit the experience?
6. **Presence** — Does the narrator, character, or argumentative consciousness feel credible and alive?
7. **Particularity** — Does the language offer something concrete, memorable, or freshly perceived?
8. **Clarity without flattening** — Is it easy to inhabit without explaining away tension, ambiguity, or texture?
9. **Earned culmination** — Does the ending feel reached rather than appended or summarized?
10. **Editorial wholeness** — Does it feel shaped as a whole rather than polished line by line?

Give one concise reader-centered reason for every `NO`. The answers are a profile, not a percentage or weighted score.

## Choose between raw and edited

After reading each independently, answer:

1. Would I rather read the edit than the raw draft?
2. Would I choose the edit as the finished piece?
3. Did the edit preserve or deepen what was most alive in the raw draft?

Use one verdict:

- **SELECT EDIT** — the covenant passes and all three answers are `YES`.
- **SIDEGRADE** — the covenant passes but preference is mixed or genuinely tied.
- **KEEP RAW** — the covenant passes and both preference questions are `NO`.
- **DIFFERENT WORK** — the covenant fails.

Do not average away disagreement between readers. Record a split as a split. Word count is descriptive only and never a quality score.