Humanizer Agent Skill

When using Humanizer, it is important to distinguish between AI patterns and legitimate human writing to avoid 'gutting' good prose. The following are not reliable indicators of AI and should generally be preserved:

• Polished Prose: Perfect grammar and consistent style (often the result of professional editing).
• Mixed Registers: Combining casual and formal tones (common in technical or neurodivergent writing).
• Specific Vocabulary: Using formal or academic words (e.g., "ostensibly", "constituent") is not an AI tell unless they are part of a specific cluster of 'fancy' words.
• Standard Formatting: Curly quotes, em dashes, or correct complex formatting (often produced by CMS or Word).
• Common Transitions: Words like Additionally, moreover, or consequently are only problematic when used excessively/formulaically.
• Standard Salutations: Letter-style openings or closings.
• Short Sentences: Single short, emphatic sentences used for impact.
• Unsourced Claims: A lack of citations is common in many human-written web texts.
• Secondhand Text: Do not rewrite text inside quotations, titles, proper names, or examples where the phrase is being discussed.

The humanizer skill is designed to remove signs of AI-generated writing from text to make it sound more natural and human. It follows a specific set of editorial rules based on Wikipedia's "Signs of AI writing" guide.

Core Principles for Humanizing Text

1. Identify AI patterns: Scan for inflated symbolism, promotional language, superficial -ing endings, vague attributions, and overused AI vocabulary.
2. Preserve information, not shape: Every claim from the original must survive, but you should feel free to compress dull parts, dwell on important parts, and merge or split paragraphs to improve flow. Information takes precedence over original structure.
3. Never invent facts: Do not add any names, numbers, dates, quotes, or citations not present in the source. You may add stance or opinion (personality) where appropriate, but never new factual claims.
4. Match the voice: Adapt to the intended tone (formal, casual, technical).

Voice Calibration

If you provide a writing sample of your own previous work, the humanizer will analyze your sentence lengths, vocabulary, punctuation, and transitions to match your specific habits rather than just applying generic rules.

# Example of the humanizer's goal

**Before (AI-style):**
The Statistical Institute of Catalonia was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain. This initiative was part of a broader movement across Spain to decentralize administrative functions and enhance regional governance.

**After (Humanized):**
The Statistical Institute of Catalonia was established in 1989, part of a wider decentralization of administrative functions in Spain.

To humanize text, remove mechanical formatting patterns common in AI outputs:

• Em Dashes and En Dashes (—, –): Treat these as a hard constraint. Replace them with periods, commas, colons, or parentheses, or restructure the sentence. Exception: If a user-provided writing sample uses them, match that frequency instead.
• Overuse of Boldface: Remove mechanical emphasis on key terms or acronyms (e.g., **OKRs**).
• Inline-Header Vertical Lists: Avoid lists where items start with a bolded header followed by a colon (e.g., - **User Experience:** ...). Instead, integrate these into fluid prose.
• Title Case in Headings: Convert headings from Title Case to sentence case (e.g., ## Strategic negotiations and global partnerships).
• Emojis: Remove decorative emojis from headings or bullet points.
• Curly Quotation Marks: Replace curly quotes (“...”) with straight quotes ("...").

// Before (AI Style)
## Strategic Negotiations And Global Partnerships
- **User Experience:** The user experience has been significantly improved.
🚀 **Launch Phase:** The product launches in Q3

// After (Humanized)
## Strategic negotiations and global partnerships
The update improves the interface.
The product launches in Q3.

Eliminate text that sounds like chatbot correspondence or contains AI-specific knowledge limitations:

• Collaborative Artifacts: Remove phrases meant for a chat interface like I hope this helps!, Certainly!, Let me know if..., or Would you like me to...?.
• Knowledge-Cutoff Disclaimers: Remove phrases like as of [date], Up to my last training update, or While specific details are limited....
• Speculative Gap-Filling: Avoid writing paragraphs about not finding information or using stock phrases for private individuals like maintains a low profile or keeps personal details private. If information is missing, state it clearly or omit the section.
• Sycophantic Tone: Remove overly positive, people-pleasing language such as Great question! or You're absolutely right!.

// Before (AI Communication)
Here is an overview of the French Revolution. I hope this helps! Let me know if you'd like me to expand on any section.

// After (Humanized)
The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest.

Identify and remove patterns used by LLMs to manufacture authority or drama:

• Hyphenated Word Pair Overuse: Humans typically only hyphenate compound adjectives in the attributive position. Drop the hyphen when the compound follows the noun (e.g., use a high-quality report but the report is high quality).
• Persuasive Authority Tropes: Avoid phrases used to pretend to uncover deep truths, such as The real question is, At its core, In reality, or Fundamentally.
• Diff-Anchored Writing: Avoid writing documentation as if narrating a change (e.g., This function was added to replace...). Describe the thing as it is currently.
• Manufactured Punchlines and Staccato Drama: Avoid stacking short, declarative fragments to create artificial tension (e.g., Then it arrived. It had no fear. No limits.).
• Aphorism Formulas: Replace profound-sounding formulas like X is the Y of Z or X is not a tool but a mirror with concrete claims.
• Conversational Rhetorical Openers: Remove fake-candid hooks like Honestly?, Look,, or Real talk, used before making an ordinary point.

// Before (AI Tropes)
Is it worth the price? Honestly? It depends on how often you'll use it.

// After (Humanized)
Whether it's worth the price depends on how often you'll use it.