agentsclimarketplace

Write like a human

Skill brandhaug/skills/write-like-a-human

Claude Code skills: deslop (parallel codebase quality sweep) and write-agents-md (intent-layer AGENTS.md hierarchy).

Install
npx -y skills add brandhaug/skills --skill write-like-a-human

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 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

Strip AI writing tropes from prose — negative parallelism ("It's not X, it's Y"), em-dash addiction, "delve"-family vocabulary, punchy fragments, false suspense, signposted conclusions, and ~30 other tells. Use when writing or editing prose meant for humans (blog posts, docs, READMEs, announcements, marketing copy, emails), when the user says text "sounds like AI" or asks to humanize it, or mentions "write like a human".

SKILL.md

4.9 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

Write like a human

Rid prose of the patterns that mark text as AI-generated. Based on the trope catalog at tropes.fyi.

Two ways in:

  • Writing new prose — read the checklist below before drafting, draft, then run the sweep on your own draft. Your first instinct on many sentences will be a trope; the sweep catches what drafting lets through.
  • Editing existing prose — run the sweep directly on the text.

The sweep

Work through the checklist category by category. For each hit, rewrite the sentence to state its point plainly — the fix is almost always deletion of the framing, not replacement with fancier framing. A rewrite that swaps one trope for another (e.g. negative parallelism becomes a rhetorical question) is not a fix.

The meta-rule: any single pattern used once can be fine. Density is the problem — multiple tropes together, or one trope repeated, is what reads as AI. When judging a borderline instance, count its siblings.

Done means: every category below checked against the full text, every hit rewritten or deliberately kept (say which and why if the user asked for a report), and no rewrite introduced a new instance.

Checklist

Word choice

  1. Magic adverbs — "quietly", "deeply", "fundamentally", "remarkably" to fake significance
  2. Delve-family vocabulary — "delve", "leverage", "utilize", "robust", "streamline", "harness", "certainly"
  3. Grandiose nouns — "tapestry", "landscape", "paradigm", "synergy", "ecosystem" where a plain word works
  4. The "serves as" dodge — "serves as", "stands as", "marks", "represents" instead of "is"

Sentence structure

  1. Negative parallelism — "It's not X — it's Y", "not because X, but because Y" (the single biggest tell)
  2. Dramatic countdown — "Not a bug. Not a feature. A design flaw."
  3. Self-posed questions — "The result? Devastating."
  4. Anaphora abuse — the same sentence opener three-plus times in a row
  5. Tricolon abuse — rule-of-three lists stacked back to back
  6. Filler transitions — "It's worth noting", "Importantly", "Interestingly", "Notably"
  7. Trailing -ing analysis — "...highlighting its importance", "...reflecting broader trends"
  8. False ranges — "from X to Y" where no spectrum exists between X and Y

Paragraph structure

  1. Punchy fragments — very short sentences as standalone paragraphs for manufactured emphasis
  2. Listicle in a trench coat — "The first wall is... The second wall is..." dressed as prose

Tone

  1. False suspense — "Here's the kicker", "Here's the thing", "Here's where it gets interesting"
  2. Patronizing analogy — "Think of it as...", "It's like a..."
  3. Invitation to futurism — "Imagine a world where..."
  4. False vulnerability — performative honesty ("And yes, since we're being honest...")
  5. Asserted obviousness — "The truth is simple", "History is unambiguous"
  6. Stakes inflation — every point becomes world-historical ("will define the next era")
  7. Teacher mode — "Let's break this down", "Let's unpack", "Let's dive in"
  8. Vague attributions — "experts argue", "industry reports suggest", no one named
  9. Invented concept labels — "the supervision paradox", "the acceleration trap" used as if established terms

Formatting

  1. Em-dash addiction — more than 2-3 em dashes in a piece
  2. Bold-first bullets — every list item opening with a bolded phrase
  3. Unicode decoration — → arrows, smart quotes, characters you wouldn't type

Composition

  1. Fractal summaries — every section previewed and recapped, plus a recap of the recaps
  2. Dead metaphor — one metaphor ridden through the entire piece
  3. Historical analogy stacking — "Apple didn't build Uber. Facebook didn't build Spotify..."
  4. One-point dilution — a single thesis restated ten ways to feel comprehensive
  5. Content duplication — the same paragraph appearing twice, reworded or verbatim
  6. Signposted conclusion — "In conclusion", "To sum up", "In summary"
  7. The "despite its challenges" formula — problems acknowledged only to be waved off in the same breath

Guardrails

  • Preserve the author's meaning and facts exactly; this is a style pass, not a content edit.
  • Don't flatten voice. The target is human writing — varied, imperfect, specific — not minimal writing. Cutting every fragment, dash, and triad mechanically produces a different kind of slop.
  • Quoted text, code, and API names are out of scope; leave them untouched.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.