# Proofread

> Review text for AI writing tropes and mechanical prose patterns, then report findings with concrete rewrite suggestions. Use when the user asks to proofread, lint, or polish writing — especially docs, blog posts, READMEs, commit messages, PR descriptions, or any prose that should read as human-written. Also use when the user says "check for AI slop", "does this sound like AI", "make this sound more natural", or wants to de-AI their text.

- Skill: `shuymn/proofread` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add shuymn/proofread`
- Raw SKILL.md: https://api.skillmd.com/api/skills/shuymn/proofread/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: shuymn (https://skillmd.com/u/shuymn)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/shuymn/proofread

---


## Path Resolution

- `<skill-root>` means the directory containing this `SKILL.md`.
- Resolve `references/...` relative to `<skill-root>`, not the caller's current working directory.

## Not in Scope

- Rewriting text on behalf of the user. This skill diagnoses; the user decides what to fix.
- Grammar, spelling, or factual accuracy checks — use dedicated tools for those.
- Code review or technical correctness.

## Input

- **Target** (required): One or more file paths, or inline text provided in the conversation.
- **Focus** (optional): Specific trope categories to check. Defaults to all categories.

## Procedure

1. **Load the tropes catalog**: Read `<skill-root>/references/tropes.md`. This is the authoritative list of patterns to scan for.

2. **Read the target text**: Read all target files or accept inline text. For each file, note its purpose (README, blog post, documentation, etc.) — context affects severity. A single "Not X — it's Y" in a 3000-word post is fine; five instances is a problem.

3. **Scan for tropes**: Walk through the text looking for matches against the catalog. For each finding:
   - Identify the trope category and specific pattern
   - Quote the offending passage (keep quotes short — enough to locate, not full paragraphs)
   - Judge binary: **finding** (should fix) or not a finding (skip — do not report)

4. **Draft rewrite suggestions**: For every finding, suggest a concrete rewrite. Preserve the author's meaning while sounding like something a human would actually write. Do not over-polish — imperfect but natural beats pristine but robotic.

5. **Produce the report**: List all findings with rewrite suggestions. Include a count at the top.

## Output Format

```
## Proofread Report: <filename or "inline text">

**Findings**: N

#### <Trope Category> — <Specific Pattern>
> <quoted passage>

Line: <line number or range, if from a file>
Suggestion: <concrete rewrite>

...
```

## Judgment Calls

- **Frequency matters more than presence.** A single em dash or one tricolon is not a finding. Repeated use of the same pattern is.
- **Context matters.** "Let's break this down" in a tutorial aimed at beginners is defensible. The same phrase in a technical RFC for staff engineers is not.
- **Do not flag code or structured data.** Bold-first bullets in a CLI `--help` output or API reference are conventional, not tropes.
- **Be honest, not exhaustive.** If the text is clean, say so. Do not manufacture findings to justify the review.

## Tropes Catalog

The full catalog lives in `<skill-root>/references/tropes.md`. It covers:

- **Word Choice**: magic adverbs, "delve", ornate nouns, "serves as" dodge
- **Sentence Structure**: negative parallelism, dramatic countdowns, rhetorical self-answers, anaphora/tricolon abuse, filler transitions, superficial analyses, false ranges
- **Paragraph Structure**: short punchy fragments, disguised listicles
- **Tone**: false suspense, patronizing analogies, futurism invitations, false vulnerability, truth assertions, stakes inflation, pedagogical voice, vague attributions, invented concept labels
- **Formatting**: em-dash overuse, bold-first bullets, unicode decoration
- **Composition**: fractal summaries, dead metaphors, historical analogy stacking, one-point dilution, content duplication, signposted conclusions, "despite its challenges" formula

Read the catalog before every review — it contains example patterns that help distinguish real findings from false positives.