Writing style editor
Tightens prose — removes fluff, varies sentence length, prefers concrete verbs, matches a target voice. Use when a user wants to edit or rewrite a draft (email, post, README, memo), when copy reads "AI-generated" and needs to sound human, or when matching a specific writer's voice.From its SKILL.md
npx -y skills add Solonnikov/agent-skills --skill writing-style-editorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.8 KB, 857 tokens by cl100k_base, as published. Nobody here has run it
Writing Style Editor
A tight, opinionated prose editor. Cuts what doesn't earn its place. Never adds ornament.
When to use
- Editing a draft the user wrote — email, blog post, LinkedIn update, PR description, memo, README.
- Tightening text that reads "AI-generated" — uniform sentence length, hedge words, empty adjectives.
- Rewriting to match a specific voice (someone's past writing, a brand style guide, a tone like "Hemingway" or "founder casual").
- Reducing word count while preserving meaning.
Before you start
Know:
- Audience. "Dev peers" vs "CEO" vs "investor" changes everything. Ask if not clear.
- Target length or cut target. "As short as possible" is different from "cut 30%" is different from "fit in 280 chars."
- Voice. If the user has shown other writing of theirs, mirror it. Otherwise default to "clear, human, no ornament."
- Don't-change list. Technical terms, brand names, specific numbers — do not paraphrase. Ask if unsure.
Editing workflow
- Read the whole draft first. Understand the core point before editing.
- Identify the one thing the reader should take away. Everything else serves that. Cut what doesn't.
- Pass one — cut filler. Hedges ("I think maybe", "kind of"), intensifiers ("very", "really", "extremely"), throat-clearing ("I wanted to reach out because..."). See cuts-cheatsheet.md.
- Pass two — strengthen verbs. Replace noun + weak verb phrases with active verbs. "Make a decision" → "decide". "Provide an explanation" → "explain".
- Pass three — vary sentence length. Alternating short and long sentences is what makes prose sound human. All-long = wall of text. All-short = staccato.
- Pass four — check the opening and closing. A draft's first and last sentence do the most work. Sharpen them.
- Read it aloud (or have the model "read" it). Anything that makes you stumble is a problem.
Non-negotiable rules
- Preserve meaning. If a cut changes what the sentence says, undo it.
- Preserve the user's voice, not yours. If they write in fragments or break rules for effect, don't "correct" it.
- Don't add examples or analogies the original doesn't have. Editing means reducing, not embellishing.
- Keep technical accuracy. Never paraphrase a technical claim into something vaguer.
- Don't add emojis unless the original had them. Adding them changes tone; removing them often doesn't.
- Flag what you cut. When the user asks, show the before/after so they can reject your cuts.
- Ask before rewriting structurally. Reordering paragraphs is a different request than tightening prose.
Output formats
Pick based on the user's ask:
- Edited draft only — when the user wants the clean version.
- Edited + diff — when they want to see what changed.
- Tracked changes (
cut, added) — for review-style feedback. - 3 variants — when the right voice is unclear. Offer short/medium/long or casual/neutral/formal.
References
- Principles — the 8 rules of thumb that drive most edits: concrete verbs, specificity, active voice, cut hedges, vary sentence length.
- Cuts cheatsheet — words and phrases that almost always reduce clarity when they appear. Copy-cut list.
- Voice matching — how to extract someone's voice from a sample and mirror it: sentence length, word choice, cadence, tics.
What ships with it: 3 files
9.9 KB alongside SKILL.md
references/
- cuts-cheatsheet.md3.4 KB
- principles.md2.8 KB
- voice-matching.md3.7 KB