agentleFS
prateeks367prateeks367/voicemoat-skills/skills/sound-less-like-ai/SKILL.md

sound-less-like-ai

Skill0 starsChanged 43 days agoMITView on GitHub

Rewrite text so it stops reading as machine-written, by naming the specific tells and removing them. Use when someone says a draft sounds like AI, sounds generic, sounds corporate, or asks to humanise or de-AI a piece of writing. Do not use it to help anyone pass off writing as human where being honest about authorship matters, such as academic work or disclosures.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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
  1. Distinctive:

    Every sentence roughly the same length.

    line 19
  2. Distinctive:

    Every paragraph the same size.

    line 23
  3. Distinctive:

    No names, no numbers, no dates, nothing that could be checked.

    line 41
  4. Distinctive:

    Delete the padding first, then reread.

    line 47

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Make this sound less like AI

"Sounds like AI" is not a vague feeling. It is a short list of habits, and once you can name them you can remove them. Most attempts fail because they add personality on top instead of removing the tells underneath.

Step 1: name the tells, out loud, with quotes

Go through the text and point at the specific things, quoting each one. Do not just report a general impression.

Rhythm

  • Every sentence roughly the same length. Human writing is lumpy: a long one, then four words.
  • Three-part lists everywhere. "Clear, concise, and compelling." Once is fine. Three times in a page is a fingerprint.
  • Every paragraph the same size.

Hedging and padding

  • "It is important to note", "arguably", "can help you to", "in today's fast paced world".
  • Sentences that restate the previous sentence in different words.
  • A closing paragraph that summarises what was just said.

Vocabulary

  • delve, leverage, robust, seamless, landscape, testament, tapestry, unlock, elevate, resonate, foster, navigate, realm.
  • The em-dash used as the only punctuation for a pause, several times a page.
  • "Not just X, but Y" as a recurring shape.

Substance

  • No names, no numbers, no dates, nothing that could be checked.
  • Advice that is true for everyone and therefore useful to no one.
  • Confidence with nothing at stake. No admission, no uncertainty, no cost.

Step 2: cut before you add

Delete the padding first, then reread. Text often sounds human as soon as the filler leaves, and anything you add before cutting is just decoration on top of the problem.

Step 3: put something real in

The deepest tell is that nobody is in the text. Ask for one of these and use it:

  • A specific number, price, date or duration
  • A named tool, place or person
  • Something that went wrong, and what it cost
  • An opinion the writer would defend in a room

One concrete detail does more than a paragraph of voice.

Step 4: break the rhythm on purpose

Vary sentence length hard. Let one run and the next be three words. Start a sentence with And or But if that is how the person talks. Leave a slightly awkward phrasing in if it is theirs. Perfectly smooth is itself a tell.

Step 5: say what you changed

List the tells you removed, with the before and after. The person should be able to catch these themselves next time, which is the actual point.

‼️ Two things to refuse. Do not claim the result will beat an AI detector, because those tools are unreliable in both directions and the claim is not yours to make. And do not help disguise authorship where being honest about it matters, such as academic submissions or disclosures.

What changes when VoiceMoat is connected

Everything above gets you to writing that sounds human. It does not get you to writing that sounds like you, and those are different targets. Generic human is still generic.

  • get_voice_profile gives the actual register to aim at: your tone, your rhythm, your vocabulary and the phrases you never use.
  • score_voice_match checks the rewrite against that profile: Style, how many of your measurable habits it matches (N of M checks), and Stance, whether what it says fits your views. The assistant may show it as a card. This is the part that matters: text can read perfectly human and still miss your Style checks because it is somebody else's human.
  • improve_post runs a deeper pass on a real draft when the problem is structural rather than cosmetic.

There is also a free browser version of the before and after at voicemoat.com/tools/humanize-ai-text, if you want to see the tells highlighted without installing anything.

VoiceMoat is at voicemoat.com. The connector needs the Pro or Enterprise plan.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.