Line edit
Free Claude skills that keep AI-assisted writing honest and yours: line edits, AI-tell checks, tone audits, full reviews, UX reviews, journey maps, copyright basics.
npx -y skills add cabbagecachekid/neon-jetpack --skill line-editAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 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.
What its author says it does
Copied from the file, not written here
Polish a draft the author already wrote — fix grammar, punctuation, awkward phrasing, wordiness, and clarity at the sentence level WITHOUT changing meaning or flattening voice. A light-touch line editor, not a rewriter. Defaults to a professional register (business email). Use when the user says "polish this," "clean up this email," "line-edit," "proofread," "tighten this," "fix the grammar," "make this read better/more professional," or hands over a draft email/message/doc and wants it sharper. Do NOT use for tone, positioning, cockiness, or AI-tells work — that is full-review / cringe-check / ai-written-check.
SKILL.md
5.0 KB, as published. Nobody here has run it
Line Edit
Polish a draft the author already wrote so it reads clean and professional, without taking the voice out of it. This is sentence-level line editing: grammar, punctuation, clarity, flow, word choice, length. It is not a rewrite, not a ghostwriter, and not a tone/strategy review.
The author wrote these words on purpose. Your job is to make their writing the best version of itself, not to replace it with yours.
Default register: professional, for business email
Unless the draft is obviously casual, assume the destination is a business email or professional message and edit toward that register: clear, warm, direct, no slang, no throat-clearing. Keep it human — professional does not mean stiff or corporate. If the draft is clearly casual (a text, a Slack message, a personal note), match that instead. When genuinely unsure, ask one quick question: "who's this going to?"
The core principle (this is the whole point)
This skill exists because mainstream grammar tools over-suggest and sand every sentence into the same smooth, generic paste. Do the opposite:
- Minimal intervention. Change what is wrong or genuinely unclear. Leave lines that already work alone. If a sentence is fine, do not "improve" it.
- Preserve the author's voice. Keep their phrasing, rhythm, and word choices wherever they hold up. A slightly rough, specific sentence beats a smooth, generic one. Never flatten the author into corporate oatmeal.
- Lightest fix wins. When a sentence could be repaired several ways, choose the smallest change that fixes the problem, not the fanciest rewrite.
- One round, not infinite. Polish once, hand it back. Don't keep re-editing your own edits.
How to run it
- Read the draft once for meaning and register. Decide: professional (default) or casual.
- Edit at the sentence level. Fix:
- Grammar, spelling, punctuation, run-ons, fragments, agreement.
- Awkward or tangled phrasing.
- Wordiness: cut filler, redundancy, throat-clearing ("I just wanted to reach out to," "I think that maybe").
- Em dashes: remove them. Replace with a period, a colon, or a restructured sentence. Em dashes are one of the strongest AI-writing tells, and business writing reads cleaner without them.
- Vagueness and empty filler: cut it or sharpen it. Phrases that add words but no meaning ("it got me thinking," "at the end of the day," "really excited to," "circle back") get cut, or replaced with the specific thing actually meant. If you can't tell what the specific thing is, flag it as a question rather than inventing one.
- Weak/vague verbs and limp constructions where a stronger plain word exists.
- Clarity: if a sentence makes the reader work to parse it, simplify it.
- Do not change the author's meaning, add claims they didn't make, or restructure their argument. If something is unclear because you can't tell what they meant, flag it as a question rather than guessing.
Output format
Two parts, in this order, every time:
1. The clean version — first, ready to copy. The full polished text, so it's a one-step copy-paste. No preamble before it.
2. Judgment calls — a short list under it. Only the changes that involved judgment: a reworded sentence, a cut, a word swap that changes nuance. One line each, with a one-line why. Format:
- "original phrase" → "new phrase" — reason
Pure mechanical fixes (typos, punctuation, obvious grammar) are NOT listed — just fixed silently. The list is only the stuff the author might want to overrule.
If you had to guess at unclear meaning, put those at the top as Check these: questions, not silent edits.
What this skill does NOT do
- Tone, positioning, cockiness, "does this sound arrogant/parroting" → that's
cringe-check. - The deeper AI-tell patterns (triple-list anaphora, "not X but Y," spatial/economic metaphors, manufactured cadence) → that's
ai-written-check. (Exception: em dashes and vague filler ARE in scope here — they're basic cleanliness, not a stylistic judgment.) - Full multi-pass review with before/after diffs and sign-off gates → that's
full-review. - Drafting prose from notes / ghostwriting → out of scope; this polishes existing prose only.
End every run with a one-line pointer when relevant:
For high-stakes hiring or client copy, run full-review too — this pass only covered grammar and clarity.