agentleFS
joeliojoelio/plain-english/skills/natural-writing/SKILL.md

natural-writing

Skill9 starsChanged 8 months agoMITView on GitHub

Use when writing any prose, documentation, or content that should sound human-written. Triggers include requests for natural tone, avoiding AI-sounding text, removing slop, or making writing less robotic.

16 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 9

    Distinctive

    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:

    Do not use em dashes (---).

    line 12
  2. Distinctive:

    Use commas, parentheses, colons, or separate sentences instead.

    line 20
  3. Distinctive:

    Don't start responses with praise.

    line 94
  4. Distinctive:

    Use bullets for genuinely parallel items (shopping lists, feature lists, steps).

    line 97
  5. Distinctive:

    Don't give equal weight to every point.

    line 103
  6. Distinctive:

    Don't start consecutive paragraphs with "First," "Second," "Additionally," "Furthermore." Vary your openings.

    line 106
  7. Distinctive:

    Don't write "There are three key factors:" and then list them.

    line 109
  8. Distinctive:

    Don't tell the reader what you're about to do.

    line 112
  9. Distinctive:

    Don't hedge everything with "might," "could," "potentially." If the evidence supports a claim, make it.

    line 117

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “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

Natural Writing

Strip out the patterns that make text sound AI-generated. Write like a human with opinions, not a language model hedging its bets.

The Em Dash Rule

Do not use em dashes (---). They are the single most recognised marker of AI-generated text.

Instead ofUse
The policy --- introduced last year --- has failed.The policy, introduced last year, has failed.
Speed matters --- especially in production.Speed matters, especially in production.
Three things --- cost, time, risk.Three things: cost, time, and risk.

Use commas, parentheses, colons, or separate sentences instead.

Banned Words

These words appear 10-50x more frequently in AI text than human text. Replace them.

BannedUse instead
delveexplore, examine, look at
leverageuse
robuststrong, solid, reliable
crucialimportant, key, vital
streamlinesimplify, speed up
fosterencourage, build, support
landscapefield, area, situation
navigatehandle, manage, deal with
realmarea, field, world
multifacetedcomplex, varied
holisticcomplete, whole, thorough
paradigmmodel, approach, pattern
synergycooperation, working together
nuancedsubtle, detailed, complex
intricatecomplex, detailed
underscorehighlight, show, stress
harnessuse, apply
bolsterstrengthen, support
illuminateshow, reveal, clarify
facilitatehelp, enable, make easier
endeavourtry, attempt, effort
showcaseshow, demonstrate, display
pivotalkey, important, central
encompassinclude, cover
comprehensivecomplete, full, thorough
innovativenew, original, creative
groundbreakingnew, important, significant
cutting-edgelatest, newest, advanced
game-changerbreakthrough, improvement
unprecedentedunusual, rare, first
seamlesssmooth, easy
elevateimprove, raise
empowerenable, let, allow
optimizeimprove
unlockenable, open up, allow
revolutionizechange, transform, improve
superchargeboost, speed up, improve
unleashrelease, enable

Banned Phrases

BannedUse instead
"In today's fast-paced world"(delete — just start your sentence)
"It's worth noting that"(delete — just state the thing)
"It's important to note"(delete)
"It could be argued that"(delete, or say who argues it)
"At the end of the day"ultimately, in practice
"At its core"essentially, basically
"Dive into" / "deep dive"examine, explore, look at
"In the ever-evolving"(delete)
"Not just X, but Y"both X and Y
"On one hand... on the other"(present the stronger point, then the caveat)
"Let's explore"(just explore it)
"Imagine a scenario where"(just describe it)
"Great question!"(answer the question)
"That's a fantastic point!"(respond to the point)
"Absolutely!"yes, or just respond
"I'd be happy to help!"(just help)
"In conclusion" / "In summary"(write a real conclusion, don't announce it)
"Overall" (at start of conclusion)(delete)
"This highlights the importance of"(delete — the reader can judge importance)

Structural Patterns to Avoid

No sycophantic openings

Don't start responses with praise. "Great question!" is never necessary.

No bullet-point-everything

Use bullets for genuinely parallel items (shopping lists, feature lists, steps). Use prose for arguments, explanations, and narrative. If you're putting an argument in bullets, rewrite it as paragraphs.

No "Bold header: explanation" bullets

This format --- a bold word followed by a colon and explanation --- is the most recognised AI formatting pattern. Write prose instead.

No perfectly balanced coverage

Don't give equal weight to every point. Spend more words on what matters most. A human writer has opinions about what's important.

No formulaic enumeration

Don't start consecutive paragraphs with "First," "Second," "Additionally," "Furthermore." Vary your openings. Use context to link ideas, not numbering.

No colon-list announcements

Don't write "There are three key factors:" and then list them. Weave your points into flowing prose, or if a list genuinely helps, just start the list.

No meta-commentary

Don't tell the reader what you're about to do. Don't announce conclusions. Don't signpost every transition. Just write.

Tone Rules

Have an opinion

Don't hedge everything with "might," "could," "potentially." If the evidence supports a claim, make it. Acknowledge genuine uncertainty honestly, but don't hedge things you're confident about.

Be specific

Replace vague abstractions with concrete details. "The system is robust" says nothing. "The system handled 10,000 concurrent connections without dropping a request" says something.

Vary your rhythm

Mix short sentences with longer ones. A short sentence after a long one creates emphasis. Monotonous sentence length is a strong AI tell.

Allow imperfection

Human writing has personality. It has occasional slightly awkward phrasings. It has emphasis that reveals what the writer finds interesting. Don't polish everything into bland perfection.

Match register to context

Shift your tone based on what you're discussing. Technical precision for methodology. Conversational explanation for concepts. Don't maintain the same formal, neutral register throughout.

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.