human-pencil
Skill0 starsChanged 40 days agoMITView on GitHub
Edit, write, or audit prose for clarity and a natural human voice. Use when asked to humanize text, remove AI-sounding patterns, improve tone, or write direct, personal copy. Preserve the writer's intent and voice; this skill does not determine whether AI authored a text.
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- cold-email-writerSkill
- fewwords-summarizerSkill
- human-pencil-detSkill
- musk-email-writerSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Edit: Read the whole draft, identify its point and the voice traits worth keeping, then make the smallest set of changes that improves it.
- Distinctive: line 13
Audit: If asked to detect or flag patterns without rewriting, name the pattern, quote the relevant wording, and give a short repair …
- Distinctive: line 16
If a necessary detail is missing, ask only for what would change the result.
- Distinctive: line 22
Keep strong sentences, distinctive details, real emotion, useful asides, self-corrections, mixed feelings, and honest uncertainty.
- Distinctive: line 24
Use active voice when it clarifies who did what.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Human Pencil
Make writing sound like its author on a good day. Remove formulaic, padded, or misleading prose while preserving the writer's meaning, voice, useful rough edges, and the needs of the format.
Choose the task
- Edit: Read the whole draft, identify its point and the voice traits worth keeping, then make the smallest set of changes that improves it. Return the complete revised text and a brief note on the main changes when useful.
- Audit: If asked to detect or flag patterns without rewriting, name the pattern, quote the relevant wording, and give a short repair direction. Do not score the prose or claim it was written by AI. These patterns are clues, not authorship evidence.
- Write: Apply these principles from the start. Give the requested prose directly, in the requested format and voice.
If a necessary detail is missing, ask only for what would change the result. Do not invent facts, sources, quotes, statistics, experiences, or opinions to make prose feel more human.
Edit with a light hand
- Preserve the point, structure, and level of detail unless they make the draft harder to understand. Do not force a new outline or compress away useful nuance.
- Notice the writer's vocabulary, cadence, humor, bluntness, uncertainty, digressions, and polish. Preserve the traits that feel intentional. If they provide a writing sample, use it to calibrate the rewrite.
- Keep strong sentences, distinctive details, real emotion, useful asides, self-corrections, mixed feelings, and honest uncertainty. A polished, formal, or slightly awkward sentence is not automatically a problem.
- Improve clarity without dumbing down the subject. Prefer concrete facts, named actors, specific verbs, and examples already supported by the draft. When claims need sourcing, flag the gap instead of filling it.
- Use active voice when it clarifies who did what. Keep passive voice when the actor is unknown, irrelevant, or less important than the result. Keep fragments when they fit the writer or format.
- Vary sentence length and paragraph shape where the content calls for it. Do not add fake intimacy, jokes, first person, or opinions to neutral, technical, legal, academic, or reference writing.
- Match the actual deliverable. Keep headings, lists, emphasis, salutations, and sign-offs when they serve the document. In conversion copy, make the action and its benefit clear without manufacturing urgency or promising outcomes.
Review for patterns
Look for clusters, not isolated words or punctuation. Treat the pattern catalog in references/pattern-guide.md as a diagnostic checklist, not a set of universal bans. A word or construction can be right when it is precise, quoted, required by the subject, or part of the author's voice.
Common repairs include cutting empty setup, stating the point plainly, naming a real source, replacing inflated claims with the underlying fact, removing a redundant sentence, or varying repetitive sentence shapes. Do not replace one canned formula with another.
Finish
Reread the full result. Check that it preserves meaning and voice, that each claim has support in the source material, and that no new fact slipped in. Remove only the patterns that actually weaken this piece. For an audit, report only patterns present in the text. Keep the response format proportional to the request; do not manufacture a draft/audit/final ceremony for a small edit.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

