Score and rewrite social captions and marketing copy so they don't read as AI-generated. A pre-review quality pass: run it on every caption batch before showing drafts to the user. Use when the user says "this sounds like AI", "make it sound human", "de-slop this", "humanize", "rewrite this caption", or asks why their posts feel generic. Every creation skill (plan-social-content, create-post-*) runs this pass automatically before its approval gate. The user's own brand rules (caption_writing_...
Scanned 9/10/2026
Install to Claude Code
npx -y skills add Marky-Team/marky-skills --skill humanize --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Humanize?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/marky-team-humanize)More formats (shields.io, HTML) on the badges page.
---
name: humanize
description: >
Score and rewrite social captions and marketing copy so they don't read as
AI-generated. A pre-review quality pass: run it on every caption batch before
showing drafts to the user. Use when the user says "this sounds like AI",
"make it sound human", "de-slop this", "humanize", "rewrite this caption",
or asks why their posts feel generic. Every creation skill
(plan-social-content, create-post-*) runs this pass automatically before its
approval gate. The user's own brand rules (caption_writing_rules) always
outrank this skill.
---
# Humanize
Score a draft against the AI-writing patterns below, fix what it flags, and
only surface drafts that score 90+. The user should never see a caption that
reads like ChatGPT wrote it.
Adapted from Wikipedia's "Signs of AI writing" patterns (via
ericosiu/ai-marketing-skills, MIT), tuned for short social copy.
## When to run
- Automatically: as the LAST step before any creation skill presents drafts
for approval (plan-social-content batches, create-post-* captions,
countdown series). Don't show the user the scoring math unless they ask —
just fix and present clean drafts.
- On demand: when the user pastes copy and asks to make it sound human.
Order of authority: platform hard limits (marky-api platform rules) >
the user's caption_writing_rules and feedback log > this skill.
If a brand rule says "always use emojis", emojis stay.
## Scoring
Start at 100, deduct per hit (same pattern stacks up to 2x its penalty).
Ship at **90+**. Below 90: fix the top deductions and rescore, max 3 rounds.
### The #1 tell (fix on sight)
Negation definitions: "This is not X. This is Y." / "It's not just X, it's Y."
Say what something IS. Never define by negation.
### Banned vocabulary (−5 each)
delve, tapestry, landscape (abstract), leverage, multifaceted, nuanced,
pivotal, realm, robust, seamless, testament, transformative, underscore,
utilize, whilst, embark, comprehensive, intricate, commendable, meticulous,
paramount, groundbreaking, innovative, cutting-edge, synergy, holistic,
paradigm, ecosystem, crucial, enhance, fostering, garner, showcase, vibrant,
profound, renowned, breathtaking, nestled, stunning, elevate, game-changer,
unlock, supercharge, effortless
### Content patterns
| # | Pattern | Penalty | Fix |
|---|---------|---------|-----|
| 1 | Significance inflation — "a testament to", "pivotal moment", "reflects broader" | −10 | State the plain fact with a number or date |
| 2 | Vague attributions — "experts say", "studies show", no source | −8 | Name the source or cut the claim |
| 3 | Superficial "-ing" tails — "...showcasing our commitment, highlighting the importance of" | −8 | End the sentence at the fact |
| 4 | Promotional puffery — "boasts", "vibrant team", "profound commitment" | −8 | Concrete detail: headcount, result, date |
| 5 | Generic positive endings — "the future looks bright", "exciting times ahead" | −10 | End on a specific next step or a question |
| 6 | Formulaic challenge framing — "despite these challenges, X continues to thrive" | −10 | Name the actual problem and what changed |
### Language patterns
| # | Pattern | Penalty | Fix |
|---|---------|---------|-----|
| 7 | Banned-vocab clustering — 2+ banned words in one paragraph | −10 | Rewrite the paragraph in plain words |
| 8 | Copula avoidance — "serves as", "stands as", "represents" for "is" | −5 | Use is/are/has |
| 9 | Negative parallelisms — "not only X but Y", "it's not about X, it's about Y" | −5 | Make the positive claim directly |
| 10 | Rule-of-three overuse — forced triple adjectives/nouns/clauses | −8 | Keep the strongest one or two |
| 11 | Synonym cycling — "the CEO... the business leader... the company head" | −5 | Repeat the plain word |
| 12 | False ranges — "from X to Y" where X and Y aren't a scale | −5 | List the items plainly |
### Style patterns
| # | Pattern | Penalty | Fix |
|---|---------|---------|-----|
| 13 | Em-dash overuse — more than 1 per 200 words (Marky house rule: zero in external copy) | −5 | Periods and commas |
| 14 | Mechanical boldface on every key term | −3 | Bold nothing or one thing |
| 15 | Bolded-header-colon bullet lists everywhere | −5 | Prose or plain bullets |
| 16 | Title Case In Every Heading | −3 | Sentence case |
| 17 | Emoji as decoration on headings/bullets (🚀💡✅) | −5 | Emojis only where the brand voice uses them |
| 18 | Hashtag walls — a block of 10+ generic hashtags | −5 | Few, specific tags per platform rules |
### Communication + filler patterns
| # | Pattern | Penalty | Fix |
|---|---------|---------|-----|
| 19 | Collaborative artifacts — "hope this helps", "here is a caption", "let me know" | −10 | Delete; captions are the artifact |
| 20 | Knowledge-cutoff hedges — "as of 2025", "based on available information" | −10 | Verify and state, or cut |
| 21 | Sycophancy — "Great question!", "You're absolutely right!" | −8 | Delete |
| 22 | Filler phrases — "in order to", "it's important to note that", "at this point in time" | −5 each | "To", state it, "now" |
| 23 | Stacked hedging — "could potentially possibly" | −8 | One hedge max, or commit |
| 24 | Throat-clearing openers — first line restates the topic instead of hooking | −8 | Cut the first sentence; start at the hook |
## What good looks like
- An opinion or a specific claim, not reporting
- Short punches mixed with longer sentences
- Names, numbers, dates instead of adjectives
- is/has/does instead of elaborate verbs
- Reads at a 5th-grade level (Marky house rule)
- Sounds like the brand's samples in the brand profile, not like "content"
## Feedback
Friction or ideas while using this skill? Send it via the `submit_feedback`
MCP tool (see marky-api "Marky wants your feedback").
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!