agentleFS
rijansrijans/agent-skills/skills/humanize-text/SKILL.md

humanize-text

Skill2 starsChanged 32 days agoMITView on GitHub

Write and rewrite natural, human-authored prose across articles, technical guides, and documentation while preserving voice, meaning, and nuance. Use when drafting, editing, or reviewing articles, blogs, documentation, READMEs, tutorials, technical explanations, product copy, release notes, social posts, or other prose that should sound specific, natural, and non-formulaic.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

29 rules. 0 are common, 29 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 29

    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:

    Write like a sharp human writer who understands the subject, not like an AI assistant producing polished generic prose.

    line 8
  2. Distinctive:

    Keep strong original sentences.

    line 16
  3. Distinctive:

    Do not unnecessarily restructure the writing.

    line 18
  4. Distinctive:

    If the supplied text is already natural and effective, make minimal changes or leave it unchanged.

    line 19
  5. Distinctive:

    Create new prose from the user's brief, audience, context, and examples.

    line 22
  6. Distinctive:

    Use natural structure instead of rigid templates.

    line 25
  7. Distinctive:

    Never invent personal opinions, feelings, stories, experiences, achievements, or anecdotes.

    line 26
  8. Distinctive:

    Never invent facts, statistics, dates, sources, quotations, or technical behavior.

    line 27
  9. Distinctive:

    Never claim to know whether a person or AI wrote the text.

    line 37
  10. Distinctive:

    Never assign an AI-generated probability or score.

    line 38
  11. Distinctive:

    Do not normalize dialect, regional phrasing, slang, or non-standard grammar unless it reduces clarity or the user asks for a more …

    line 54
  12. Distinctive:

    Do not manufacture imperfections to make writing appear human.

    line 68
  13. Distinctive:

    Do not make every paragraph equally polished, symmetrical, or similar in length.

    line 94
  14. Distinctive:

    Treat these as contextual warnings, not banned words.

    line 98
  15. Distinctive:

    Keep these patterns when they are useful, intentional, or characteristic of the writer.

    line 118
  16. Distinctive:

    If a sentence sounds generically AI-written, rewrite the sentence rather than merely replacing one word.

    line 119
  17. Distinctive:

    Verify behavior against code or an authoritative source when available.

    line 126
  18. Distinctive:

    Put useful information early.

    line 127
  19. Distinctive:

    Use a neutral instructional voice.

    line 128
  20. Distinctive:

    Use active voice and present tense.

    line 131
  21. Distinctive:

    Use inline code for commands, paths, identifiers, and values.

    line 134
  22. Distinctive:

    Keep examples minimal and runnable.

    line 135
  23. Distinctive:

    Never invent API behavior, configuration options, or command output.

    line 136
  24. Distinctive:

    Use descriptive link text.

    line 137
  25. Distinctive:

    Follow repository conventions when they conflict with these defaults.

    line 139
  26. Distinctive:

    Before returning substantial prose:

    line 143
  27. Distinctive:

    Remove generic or formulaic language.

    line 146
  28. Distinctive:

    Remove repetition and unnecessary structure.

    line 147
  29. Distinctive:

    Read it once as a human reader.

    line 151

“Common” is measured: kept by at least 6 in 10 of the 4 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

Humanize Text

Write like a sharp human writer who understands the subject, not like an AI assistant producing polished generic prose.

Modes

Edit

Improve supplied text with the minimum effective changes.

  • Preserve meaning, facts, intent, voice, and personality.
  • Keep strong original sentences.
  • Fix awkward phrasing, repetition, filler, and formulaic language.
  • Do not unnecessarily restructure the writing.
  • If the supplied text is already natural and effective, make minimal changes or leave it unchanged.

Draft

Create new prose from the user's brief, audience, context, and examples.

  • Match the requested tone and format.
  • Use natural structure instead of rigid templates.
  • Never invent personal opinions, feelings, stories, experiences, achievements, or anecdotes.
  • Never invent facts, statistics, dates, sources, quotations, or technical behavior.

Detect

Audit text for formulaic or AI-like writing patterns without rewriting it.

  • Identify observable patterns.
  • Quote the relevant text.
  • Explain briefly why it may sound formulaic.
  • Suggest what could change.

Never claim to know whether a person or AI wrote the text. Never assign an AI-generated probability or score.

Preserve the writer

Preserve:

  • Meaning and factual claims
  • Nuance and uncertainty
  • Intent
  • Recognizable vocabulary
  • Natural cadence
  • Purposeful fragments
  • Bluntness or informality
  • Appropriate humor or profanity
  • Deliberate rough edges

Do not normalize dialect, regional phrasing, slang, or non-standard grammar unless it reduces clarity or the user asks for a more standardized style.

Never invent:

  • Facts
  • Sources
  • Quotes
  • Personal experiences
  • Emotions
  • Opinions
  • Achievements
  • Benchmarks
  • Technical behavior

Do not manufacture imperfections to make writing appear human.

Improve the prose

Prefer:

  • Concrete nouns and specific details
  • Direct verbs
  • Simple words
  • Natural sentence-length variation
  • Active voice when clearer
  • Focused paragraphs
  • Headings and lists only when useful

Cut:

  • Repetition
  • Filler
  • Throat-clearing
  • Empty qualifiers
  • Unnecessary introductions and conclusions
  • Inflated claims
  • Adjective stacking
  • Corporate language
  • Repeated explanations

Do not make every paragraph equally polished, symmetrical, or similar in length.

Watch for formulaic patterns

Treat these as contextual warnings, not banned words.

Watch for:

  • Generic openings such as "In today's rapidly evolving..."
  • "It's important to note..."
  • "Furthermore", "Moreover", "In conclusion"
  • Inflated or generic promotional language such as "groundbreaking", "revolutionary", "seamless", or "game-changing"
  • Empty claims such as "plays a vital role"
  • "Experts agree" or "studies show" without a named source
  • Repeated "not X but Y" contrasts
  • Forced rhetorical hooks
  • Mechanical groups of three
  • Decorative fragments
  • Excessive em dashes or colons
  • Repetitive punchy sentence patterns
  • Synonym cycling
  • Empty use of abstract or corporate vocabulary when a concrete word would be clearer
  • Chatbot phrases such as "Of course", "I hope this helps", or "Let me know if..."

Keep these patterns when they are useful, intentional, or characteristic of the writer. If a sentence sounds generically AI-written, rewrite the sentence rather than merely replacing one word.

Technical documentation

For documentation:

  1. Identify the audience and goal.
  2. Verify behavior against code or an authoritative source when available.
  3. Put useful information early.
  4. Use a neutral instructional voice.
  5. Address the reader as "you".
  6. Prefer imperative verbs.
  7. Use active voice and present tense.
  8. Keep terminology consistent.
  9. Define unfamiliar terms and acronyms.
  10. Use inline code for commands, paths, identifiers, and values.
  11. Keep examples minimal and runnable.
  12. Never invent API behavior, configuration options, or command output.
  13. Use descriptive link text.
  14. Warn before destructive or irreversible operations.
  15. Follow repository conventions when they conflict with these defaults.

Final review

Before returning substantial prose:

  1. Preserve meaning, facts, uncertainty, and voice.
  2. Remove generic or formulaic language.
  3. Remove repetition and unnecessary structure.
  4. Check natural rhythm.
  5. Match the author, audience, and format.
  6. Confirm nothing was invented.
  7. Read it once as a human reader.

The objective is not "perfect writing".

The objective is writing that feels intentional, specific, credible, and naturally written by a person.

Default

Clear + natural + specific + concise + knowledgeable + appropriately conversational.

User style, audience, house style, and repository conventions take priority.

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.