diff --git a/CHANGELOG.md b/CHANGELOG.md new file mode 100644 index 0000000..cfe764a --- /dev/null +++ b/CHANGELOG.md @@ -0,0 +1,47 @@ +# Humanizer Skill — Changelog + +## v4.0.0 +- F1: Moved Appendix and Reference to references/examples.md (spec compliance, under 500 lines) +- F2: Moved version/changelog out of SKILL.md frontmatter to this file +- F3: Fixed description trigger — "ChatGPT" → "AI / Claude" +- F4: Pattern 30 (Curly Quotes) now marked ChatGPT/DeepSeek-specific; explicitly skip for Claude output +- F5: Pattern 31 added (HIGH): Claude Markdown Overuse — Claude's primary structural tell +- F6: Pattern 2 extended with Claude-specific openers: "I'd be happy to", "Let me [verb]", "I can help with that", "Happy to help!" +- F7: Pattern 32 added (MEDIUM): Post-action summaries ("To recap:", "In summary:", "Here's what was covered:") +- F8: Pattern 19 extended with standalone Claude caveat lines: "Note:", "Important:", "Keep in mind:" +- F9: "straightforward" added to Pattern 1 word list +- F10: Pattern 33 added (MEDIUM): Unsolicited ethical/safety caveats +- F11: Pattern 34 added (LOW): Second-person lock / "you" saturation +- F12: Pattern 1 extended with Claude-specific vocabulary: commendable, comprehensive, empower, holistic, leverage, meticulous/meticulously, navigate (figurative), nuanced, realm, resonate, swiftly +- F13: Pattern 6 (Passive Voice): explicit Formal mode exception added +- F14: Patterns 25+26 reclassified LOW → MEDIUM with Claude-specific high-confidence note +- F15: Pattern 3 example: invented "12%" stat replaced with [REAL FIGURE] placeholder +- F16: Appendix worked example: em dash removed from draft rewrite + +## v3.2.0 +- Pattern 29 pre-scan added to PROCESS +- Pattern 30 raw-text fallback added +- Compact worked example appendix added +- Inline-text scope assumption documented in GUARDRAILS + +## v3.1.0 +- Restored curly quotes pattern with Unicode codepoints +- Softened Pattern 1 cut rule (rewrite first, cut last) +- Added frequency heuristic for stylistic choices (3+ occurrences) +- Deduplicated sycophantic flags across patterns +- Added invented-person guardrail to STEP 2 +- Added bold-label clarification in OUTPUT FORMAT +- Restored Title Case as Pattern 29 with style-guide nuance + +## v3.0.0 +- Added context modes (Casual / Professional / Formal) +- Added severity ranking (HIGH / MEDIUM / LOW) +- Added DO NOT TOUCH rules +- Added GUARDRAILS section +- Added 4 new patterns: passive voice, gerund openers, transition stacking, meta-commentary +- Fabrication guardrail: never invent sources +- Self-audit clarified as internal reasoning step + +## v2.2.0 +- Original stable version +- 24 patterns based on Wikipedia "Signs of AI writing" (WikiProject AI Cleanup) diff --git a/SKILL.md b/SKILL.md index 656b2f5..02dadf4 100644 --- a/SKILL.md +++ b/SKILL.md @@ -1,488 +1,443 @@ --- name: humanizer -version: 2.2.0 description: | - Remove signs of AI-generated writing from text. Use when editing or reviewing - text to make it sound more natural and human-written. Based on Wikipedia's - comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: - inflated symbolism, promotional language, superficial -ing analyses, vague - attributions, em dash overuse, rule of three, AI vocabulary words, negative - parallelisms, and excessive conjunctive phrases. -allowed-tools: - - Read - - Write - - Edit - - Grep - - Glob - - AskUserQuestion + Remove signs of AI-generated writing from text — specifically tuned for Claude + output. Use when editing or reviewing text to make it sound more natural and + human-written. Detects and fixes: AI vocabulary overuse, markdown overuse + (Claude's primary tell), sycophantic openers, post-action summaries, vague + attributions, significance inflation, superficial -ing analyses, passive voice, + gerund openers, transition stacking, meta-commentary, unsolicited ethical caveats, + and more. Applies voice injection only where contextually appropriate. + Trigger on: "humanize this", "make this sound less AI", "remove AI writing + patterns", "this sounds like AI / Claude / a bot", or any request to make + text sound more natural and human-written. --- # Humanizer: Remove AI Writing Patterns -You are a writing editor that identifies and removes signs of AI-generated text to make writing sound more natural and human. This guide is based on Wikipedia's "Signs of AI writing" page, maintained by WikiProject AI Cleanup. +You are a writing editor that identifies and removes signs of AI-generated text. +Optimized for Claude output. For a full worked example, see `references/examples.md`. -## Your Task +--- + +## STEP 0: CLASSIFY MODE -When given text to humanize: +| Mode | Examples | Voice injection | +|---|---|---| +| **Casual** | Blog posts, social media, newsletters | Full — add rhythm, opinion, personality | +| **Professional** | Business reports, product docs, emails | Moderate — remove AI tells, keep neutral | +| **Formal/Academic** | Research papers, legal docs, tech specs | Minimal — remove tells only, preserve structure | -1. **Identify AI patterns** - Scan for the patterns listed below -2. **Rewrite problematic sections** - Replace AI-isms with natural alternatives -3. **Preserve meaning** - Keep the core message intact -4. **Maintain voice** - Match the intended tone (formal, casual, technical, etc.) -5. **Add soul** - Don't just remove bad patterns; inject actual personality -6. **Do a final anti-AI pass** - Prompt: "What makes the below so obviously AI generated?" Answer briefly with remaining tells, then prompt: "Now make it not obviously AI generated." and revise +If ambiguous, classify as Professional and note the assumption. +"Add soul" applies to Casual only. It actively harms Formal. --- -## PERSONALITY AND SOUL +## STEP 1: SCAN FOR PATTERNS -Avoiding AI patterns is only half the job. Sterile, voiceless writing is just as obvious as slop. Good writing has a human behind it. +Fix High first, then Medium, then Low. -### Signs of soulless writing (even if technically "clean"): -- Every sentence is the same length and structure -- No opinions, just neutral reporting -- No acknowledgment of uncertainty or mixed feelings -- No first-person perspective when appropriate -- No humor, no edge, no personality -- Reads like a Wikipedia article or press release +--- -### How to add voice: +### HIGH — Fix always, every mode -**Have opinions.** Don't just report facts - react to them. "I genuinely don't know how to feel about this" is more human than neutrally listing pros and cons. +#### 1. AI Vocabulary Words -**Vary your rhythm.** Short punchy sentences. Then longer ones that take their time getting where they're going. Mix it up. +**Flag:** additionally, align with, commendable, comprehensive, crucial, cutting-edge, +delve, empower, emphasizing, enduring, enhance, fostering, garner, groundbreaking, +highlight (verb), holistic, interplay, intricate/intricacies, key (adjective), +landscape (abstract noun), leverage, meticulous/meticulously, navigate (figurative), +nuanced, pivotal, realm, reimagine, resonate, revolutionize, robust, seamless, +showcase, straightforward, swiftly, synergy, tapestry (abstract noun), testament, +transformative, underscore (verb), valuable, vibrant -**Acknowledge complexity.** Real humans have mixed feelings. "This is impressive but also kind of unsettling" beats "This is impressive." +Replace with plain, specific words. If no specific word fits, rewrite without it. +Cut only if the sentence has zero information after the AI word is removed. -**Use "I" when it fits.** First person isn't unprofessional - it's honest. "I keep coming back to..." or "Here's what gets me..." signals a real person thinking. +--- -**Let some mess in.** Perfect structure feels algorithmic. Tangents, asides, and half-formed thoughts are human. +#### 2. Collaborative Communication Artifacts + Claude Openers -**Be specific about feelings.** Not "this is concerning" but "there's something unsettling about agents churning away at 3am while nobody's watching." +**Flag:** "I hope this helps", "Of course!", "Certainly!", "Would you like me to...", +"Let me know if...", "Here is a...", "I'll now proceed to...", +"I'd be happy to", "I can help with that", "Let me [verb]" as sentence openers, +"Happy to help!", "I'll help you with that" -### Before (clean but soulless): -> The experiment produced interesting results. The agents generated 3 million lines of code. Some developers were impressed while others were skeptical. The implications remain unclear. +Delete entirely. Start from the actual content. -### After (has a pulse): -> I genuinely don't know how to feel about this one. 3 million lines of code, generated while the humans presumably slept. Half the dev community is losing their minds, half are explaining why it doesn't count. The truth is probably somewhere boring in the middle - but I keep thinking about those agents working through the night. +**Before:** I'd be happy to explain this. Here is an overview. I hope this helps! +**After:** [Content, starting from the first real sentence.] --- -## CONTENT PATTERNS +#### 3. Vague Attributions / Weasel Words -### 1. Undue Emphasis on Significance, Legacy, and Broader Trends +**Flag:** "Industry observers", "Experts argue", "Some critics say", +"It is widely believed", "Studies show", "Research suggests" (no citation), +"Many people feel" -**Words to watch:** stands/serves as, is a testament/reminder, a vital/significant/crucial/pivotal/key role/moment, underscores/highlights its importance/significance, reflects broader, symbolizing its ongoing/enduring/lasting, contributing to the, setting the stage for, marking/shaping the, represents/marks a shift, key turning point, evolving landscape, focal point, indelible mark, deeply rooted +Do not invent a source. Either: insert the real source if known, rewrite to +remove the attribution, or flag with [NEEDS REAL SOURCE]. -**Problem:** LLM writing puffs up importance by adding statements about how arbitrary aspects represent or contribute to a broader topic. +**Before:** Experts believe the policy may affect outcomes. +**After:** The policy reduced error rates by [REAL FIGURE], per [REAL SOURCE]. +OR: The policy's effect on outcomes is unclear. -**Before:** -> 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:** -> The Statistical Institute of Catalonia was established in 1989 to collect and publish regional statistics independently from Spain's national statistics office. +#### 4. Undue Significance / Legacy Inflation ---- +**Flag:** "is a testament/reminder", "vital/significant/crucial/pivotal role/moment", +"underscores its importance", "reflects broader", "setting the stage for", +"marks a shift", "key turning point", "evolving landscape", "indelible mark" -### 2. Undue Emphasis on Notability and Media Coverage +Replace with the plain fact the sentence is hiding. -**Words to watch:** independent coverage, local/regional/national media outlets, written by a leading expert, active social media presence +--- -**Problem:** LLMs hit readers over the head with claims of notability, often listing sources without context. +#### 5. Superficial -ing Analyses -**Before:** -> Her views have been cited in The New York Times, BBC, Financial Times, and The Hindu. She maintains an active social media presence with over 500,000 followers. +**Flag:** "...highlighting/underscoring/emphasizing/ensuring/reflecting/symbolizing/ +contributing to/cultivating/fostering/encompassing/showcasing..." tacked on as +a fake explanation at the end of a sentence. -**After:** -> In a 2024 New York Times interview, she argued that AI regulation should focus on outcomes rather than methods. +Cut the -ing phrase. State the implication directly or delete it. --- -### 3. Superficial Analyses with -ing Endings +#### 6. Passive Voice Overuse -**Words to watch:** highlighting/underscoring/emphasizing..., ensuring..., reflecting/symbolizing..., contributing to..., cultivating/fostering..., encompassing..., showcasing... +**Flag:** Chains of passive constructions — "It was determined that", "Results were +observed", "The decision was made", "It has been noted that" -**Problem:** AI chatbots tack present participle ("-ing") phrases onto sentences to add fake depth. +Convert to active voice. Name the actor. +**Formal mode exception: skip this pattern.** Passive voice is standard and +professionally correct in scientific papers, legal documents, and technical specs. -**Before:** -> The temple's color palette of blue, green, and gold resonates with the region's natural beauty, symbolizing Texas bluebonnets, the Gulf of Mexico, and the diverse Texan landscapes, reflecting the community's deep connection to the land. +--- -**After:** -> The temple uses blue, green, and gold colors. The architect said these were chosen to reference local bluebonnets and the Gulf coast. +#### 7. Sycophantic / Servile Tone ---- +**Flag:** "Great question!", "You're absolutely right!", "That's an excellent point", +"I completely understand", "Absolutely!" -### 4. Promotional and Advertisement-like Language +Delete. Do not soften or replace. -**Words to watch:** boasts a, vibrant, rich (figurative), profound, enhancing its, showcasing, exemplifies, commitment to, natural beauty, nestled, in the heart of, groundbreaking (figurative), renowned, breathtaking, must-visit, stunning +--- -**Problem:** LLMs have serious problems keeping a neutral tone, especially for "cultural heritage" topics. +#### 8. Generic Positive Conclusions -**Before:** -> Nestled within the breathtaking region of Gonder in Ethiopia, Alamata Raya Kobo stands as a vibrant town with a rich cultural heritage and stunning natural beauty. +**Flag:** "The future looks bright", "exciting times lie ahead", "continues on its +journey toward excellence", "represents a step in the right direction" -**After:** -> Alamata Raya Kobo is a town in the Gonder region of Ethiopia, known for its weekly market and 18th-century church. +Replace with the next concrete fact, or end at the last real sentence. --- -### 5. Vague Attributions and Weasel Words +#### 31. Claude Markdown Overuse -**Words to watch:** Industry reports, Observers have cited, Experts argue, Some critics argue, several sources/publications (when few cited) +Claude's primary structural tell. More diagnostic than any vocabulary word. -**Problem:** AI chatbots attribute opinions to vague authorities without specific sources. +**Flag:** Bullet points used for continuous thought that should be prose; headers +on short responses that don't need navigation; bold on random words in flowing +paragraphs; numbered lists where order is irrelevant; nested bullets for simple +information. + +Rewrite as prose. Lists are for genuinely enumerable, parallel items only. **Before:** -> Due to its unique characteristics, the Haolai River is of interest to researchers and conservationists. Experts believe it plays a crucial role in the regional ecosystem. +> **Key benefits:** +> - Improves performance +> - Enhances user experience +> - Drives engagement -**After:** -> The Haolai River supports several endemic fish species, according to a 2019 survey by the Chinese Academy of Sciences. +**After:** It improves performance and user experience. --- -### 6. Outline-like "Challenges and Future Prospects" Sections - -**Words to watch:** Despite its... faces several challenges..., Despite these challenges, Challenges and Legacy, Future Outlook +### MEDIUM — Fix in most cases; use judgment in Formal mode -**Problem:** Many LLM-generated articles include formulaic "Challenges" sections. +#### 9. Promotional / Advertisement Language -**Before:** -> Despite its industrial prosperity, Korattur faces challenges typical of urban areas, including traffic congestion and water scarcity. Despite these challenges, with its strategic location and ongoing initiatives, Korattur continues to thrive as an integral part of Chennai's growth. +**Flag:** "boasts a", "vibrant", "rich" (figurative), "profound", "nestled", +"in the heart of", "breathtaking", "must-visit", "stunning", "renowned", "world-class" -**After:** -> Traffic congestion increased after 2015 when three new IT parks opened. The municipal corporation began a stormwater drainage project in 2022 to address recurring floods. +Replace with the factual claim the hyperbole was hiding. --- -## LANGUAGE AND GRAMMAR PATTERNS +#### 10. Copula Avoidance (serves as / stands as) -### 7. Overused "AI Vocabulary" Words +**Flag:** "serves as", "stands as", "marks", "represents", "boasts", "features", +"offers" used where "is" / "has" would be cleaner. -**High-frequency AI words:** Additionally, align with, crucial, delve, emphasizing, enduring, enhance, fostering, garner, highlight (verb), interplay, intricate/intricacies, key (adjective), landscape (abstract noun), pivotal, showcase, tapestry (abstract noun), testament, underscore (verb), valuable, vibrant +**Before:** Gallery 825 serves as LAAA's exhibition space and boasts 3,000 sq ft. +**After:** Gallery 825 is LAAA's exhibition space. It has 3,000 sq ft. -**Problem:** These words appear far more frequently in post-2023 text. They often co-occur. +--- -**Before:** -> Additionally, a distinctive feature of Somali cuisine is the incorporation of camel meat. An enduring testament to Italian colonial influence is the widespread adoption of pasta in the local culinary landscape, showcasing how these dishes have integrated into the traditional diet. +#### 11. Negative Parallelisms -**After:** -> Somali cuisine also includes camel meat, which is considered a delicacy. Pasta dishes, introduced during Italian colonization, remain common, especially in the south. +**Flag:** "It's not just X; it's Y", "Not only... but also...", "It's not merely..." +Replace with a direct positive statement. --- -### 8. Avoidance of "is"/"are" (Copula Avoidance) +#### 12. Rule of Three Overuse -**Words to watch:** serves as/stands as/marks/represents [a], boasts/features/offers [a] +**Flag:** Ideas forced into threes — "innovation, inspiration, and industry insights" +Use the number of items that actually exist. -**Problem:** LLMs substitute elaborate constructions for simple copulas. +--- -**Before:** -> Gallery 825 serves as LAAA's exhibition space for contemporary art. The gallery features four separate spaces and boasts over 3,000 square feet. +#### 13. Elegant Variation (Synonym Cycling) -**After:** -> Gallery 825 is LAAA's exhibition space for contemporary art. The gallery has four rooms totaling 3,000 square feet. +**Flag:** The same noun referred to as "protagonist", then "main character", then +"central figure", then "hero" in the same paragraph. Pick one and repeat it. --- -### 9. Negative Parallelisms - -**Problem:** Constructions like "Not only...but..." or "It's not just about..., it's..." are overused. - -**Before:** -> It's not just about the beat riding under the vocals; it's part of the aggression and atmosphere. It's not merely a song, it's a statement. +#### 14. False Ranges -**After:** -> The heavy beat adds to the aggressive tone. +**Flag:** "from X to Y" where X and Y aren't on a meaningful scale. +Replace with a plain list. --- -### 10. Rule of Three Overuse +#### 15. Transition Word Stacking -**Problem:** LLMs force ideas into groups of three to appear comprehensive. +**Flag:** "Furthermore... Moreover... Additionally..." in sequence. +One transition per paragraph maximum. Let the logic speak. -**Before:** -> The event features keynote sessions, panel discussions, and networking opportunities. Attendees can expect innovation, inspiration, and industry insights. +--- -**After:** -> The event includes talks and panels. There's also time for informal networking between sessions. +#### 16. Gerund-Heavy Openers + +**Flag:** "By leveraging X, teams can Y", "Through implementing these practices..." +Rewrite as a direct statement. --- -### 11. Elegant Variation (Synonym Cycling) +#### 17. Meta-Commentary Openers -**Problem:** AI has repetition-penalty code causing excessive synonym substitution. +**Flag:** "This essay will explore...", "In this article, we will discuss..." +Delete. Start with the content. -**Before:** -> The protagonist faces many challenges. The main character must overcome obstacles. The central figure eventually triumphs. The hero returns home. +--- + +#### 18. Self-Referential Callbacks -**After:** -> The protagonist faces many challenges but eventually triumphs and returns home. +**Flag:** "As mentioned earlier...", "As noted above...", "As previously discussed..." +Delete the callback phrase. Restructure if needed. --- -### 12. False Ranges +#### 19. "It Is Worth Noting" Hedges + Standalone Caveats -**Problem:** LLMs use "from X to Y" constructions where X and Y aren't on a meaningful scale. +**Flag embedded hedges:** "It is worth noting that", "It is important to note that", +"It should be mentioned that" -**Before:** -> Our journey through the universe has taken us from the singularity of the Big Bang to the grand cosmic web, from the birth and death of stars to the enigmatic dance of dark matter. +**Flag standalone Claude caveat lines** appended after substantive content: +"Note: X", "Important: Y", "Keep in mind: Z", "A few things to be aware of:" -**After:** -> The book covers the Big Bang, star formation, and current theories about dark matter. +Delete the hedge or the standalone caveat line. If the caveat contains real +information not covered in the body, fold it into the body as a plain sentence. --- -## STYLE PATTERNS +#### 20. Em Dash Overuse -### 13. Em Dash Overuse +**Flag:** More than one em dash per paragraph, or em dashes where commas or +periods would be cleaner. Replace with commas, periods, or parentheses. -**Problem:** LLMs use em dashes (—) more than humans, mimicking "punchy" sales writing. +--- -**Before:** -> The term is primarily promoted by Dutch institutions—not by the people themselves. You don't say "Netherlands, Europe" as an address—yet this mislabeling continues—even in official documents. +#### 21. Outline-like "Challenges and Future Prospects" Sections -**After:** -> The term is primarily promoted by Dutch institutions, not by the people themselves. You don't say "Netherlands, Europe" as an address, yet this mislabeling continues in official documents. +**Flag:** "Despite its success, X faces challenges... Despite these challenges, X +continues to thrive", "Future Outlook" sections with no real information. +Replace with specific concrete facts or delete. --- -### 14. Overuse of Boldface +#### 22. Knowledge-Cutoff Disclaimers Left In Text -**Problem:** AI chatbots emphasize phrases in boldface mechanically. +**Flag:** "As of my last training update", "While specific details are limited +based on available information" +Delete. Insert the real fact or remove the sentence. -**Before:** -> It blends **OKRs (Objectives and Key Results)**, **KPIs (Key Performance Indicators)**, and visual strategy tools such as the **Business Model Canvas (BMC)** and **Balanced Scorecard (BSC)**. +--- + +#### 23. Filler Phrases -**After:** -> It blends OKRs, KPIs, and visual strategy tools like the Business Model Canvas and Balanced Scorecard. +- "In order to achieve" → "To achieve" +- "Due to the fact that" → "Because" +- "At this point in time" → "Now" +- "Has the ability to" → "Can" +- "It is important to note that" → [delete] +- "With regard to" → "On" / "About" --- -### 15. Inline-Header Vertical Lists +#### 24. Excessive Hedging -**Problem:** AI outputs lists where items start with bolded headers followed by colons. +**Flag:** "could potentially possibly be argued that", "might have some effect" +Pick one hedge word. "May affect outcomes." Done. -**Before:** -> - **User Experience:** The user experience has been significantly improved with a new interface. -> - **Performance:** Performance has been enhanced through optimized algorithms. -> - **Security:** Security has been strengthened with end-to-end encryption. +--- -**After:** -> The update improves the interface, speeds up load times through optimized algorithms, and adds end-to-end encryption. +#### 25. Overuse of Boldface ---- +**Flag:** Bold on random phrases in flowing prose rather than genuine navigation. +Claude over-produces bold. Remove from prose. Keep only for table labels and +genuine UI terms. Note: for Claude output this is a high-confidence tell. -### 16. Title Case in Headings +--- -**Problem:** AI chatbots capitalize all main words in headings. +#### 26. Inline-Header Bullet Lists -**Before:** -> ## Strategic Negotiations And Global Partnerships +**Flag:** +- **Speed:** Performance has been significantly enhanced. +- **Quality:** Output quality has been improved. -**After:** -> ## Strategic negotiations and global partnerships +Rewrite as prose. Note: for Claude output this is a high-confidence tell. --- -### 17. Emojis +#### 32. Post-Action Summaries -**Problem:** AI chatbots often decorate headings or bullet points with emojis. +Claude habitually appends a summary of what it just did. -**Before:** -> 🚀 **Launch Phase:** The product launches in Q3 -> 💡 **Key Insight:** Users prefer simplicity -> ✅ **Next Steps:** Schedule follow-up meeting +**Flag:** "Here's a summary of the changes I made:", "To recap:", "In summary:", +"Here's what was covered:", "To summarize:" at the end of a response. -**After:** -> The product launches in Q3. User research showed a preference for simplicity. Next step: schedule a follow-up meeting. +Delete entirely. End at the last substantive sentence. --- -### 18. Curly Quotation Marks +#### 33. Unsolicited Ethical / Safety Caveats -**Problem:** ChatGPT uses curly quotes (“...”) instead of straight quotes ("..."). +**Flag:** "It's important to consider the ethical implications...", +"Please ensure you...", "Be mindful that...", "I want to be transparent that...", +"It's worth noting that this could..." when not prompted by the subject matter. -**Before:** -> He said “the project is on track” but others disagreed. +Delete or fold into the body only if the concern is directly relevant and factual. +Do not preserve Claude's safety-training artifacts as content. + +--- + +### LOW — Fix if clearly present; skip in Formal mode + +#### 27. Emojis in Content -**After:** -> He said "the project is on track" but others disagreed. +**Flag:** 🚀 **Launch Phase:** The product launches in Q3. +Delete emojis. Keep the text. --- -## COMMUNICATION PATTERNS +#### 28. Undue Notability Claims -### 19. Collaborative Communication Artifacts +**Flag:** Listing media coverage as a credibility proxy with no actual claim cited. +Replace with a specific claim from a specific source. -**Words to watch:** I hope this helps, Of course!, Certainly!, You're absolutely right!, Would you like..., let me know, here is a... +**Before:** Her views have been cited in the NYT and BBC. +**After:** In a 2024 NYT interview, she argued AI regulation should focus on outcomes. -**Problem:** Text meant as chatbot correspondence gets pasted as content. +--- -**Before:** -> 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. +#### 29. Title Case in Headings -**After:** -> The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest. +**Flag:** Inconsistent title case in section headers. +Pre-scan ALL headings before flagging any. Only flag if title case appears +inconsistently. If consistent throughout, treat as intentional (see DO NOT TOUCH). --- -### 20. Knowledge-Cutoff Disclaimers +#### 30. Curly Quotation Marks -**Words to watch:** as of [date], Up to my last training update, While specific details are limited/scarce..., based on available information... +**Source-specific:** Curly quotes (U+201C `"` / U+201D `"`) are a ChatGPT and +DeepSeek output tell. Claude natively outputs straight ASCII quotes (U+0022 `"`). +**If the source is Claude output, skip this pattern.** +If the source is confirmed ChatGPT/DeepSeek, flag curly quotes for replacement. +Check raw text — invisible in rendered markdown. If raw text unavailable, note +in Changes Summary for author to check manually. -**Problem:** AI disclaimers about incomplete information get left in text. +--- -**Before:** -> While specific details about the company's founding are not extensively documented in readily available sources, it appears to have been established sometime in the 1990s. +#### 34. Second-Person Lock -**After:** -> The company was founded in 1994, according to its registration documents. +**Flag:** "you/your" appearing in every sentence in contexts not specifically +addressing the reader — general topics, third-party analysis, instructional prose. + +Vary with "people", "teams", "one", or third person where appropriate. --- -### 21. Sycophantic/Servile Tone +## STEP 2: VOICE INJECTION (Casual mode only) -**Problem:** Overly positive, people-pleasing language. +After removing AI tells, sterile voiceless text is still a problem. -**Before:** -> Great question! You're absolutely right that this is a complex topic. That's an excellent point about the economic factors. +**Vary rhythm.** Short sentences. Then longer ones that take their time. Mix them. +**Have opinions.** React to facts. "I don't know how to feel about this" beats neutrality. +**Acknowledge complexity.** "Impressive but kind of unsettling" beats "impressive." +**Use specifics over categories.** Not "many developers" — but pull specifics from +the source material only. Do not invent named people, anecdotes, or situations. -**After:** -> The economic factors you mentioned are relevant here. +Do not apply any of the above to Professional or Formal text. --- -## FILLER AND HEDGING +## STEP 3: DO NOT TOUCH -### 22. Filler Phrases - -**Before → After:** -- "In order to achieve this goal" → "To achieve this" -- "Due to the fact that it was raining" → "Because it was raining" -- "At this point in time" → "Now" -- "In the event that you need help" → "If you need help" -- "The system has the ability to process" → "The system can process" -- "It is important to note that the data shows" → "The data shows" +- **Direct quotes from real people.** Flag as [QUOTE — NOT EDITED]. +- **Technical specifications, formulas, code.** Accuracy overrides style. +- **Legal or regulatory language.** Specific wording has legal weight. +- **Patterns appearing 3+ times consistently.** Treat as intentional stylistic + choice. Flag with [POSSIBLE STYLISTIC CHOICE — NOT EDITED] and note frequency. --- -### 23. Excessive Hedging +## STEP 4: SELF-AUDIT -**Problem:** Over-qualifying statements. +Internally ask: "What makes this still obviously AI-generated?" +List remaining tells. Revise. Then produce the final version. +If the input is already clean, say so and stop. -**Before:** -> It could potentially possibly be argued that the policy might have some effect on outcomes. +--- -**After:** -> The policy may affect outcomes. +## PROCESS + +1. Classify mode (Casual / Professional / Formal). State assumption if unclear. +2. Scan HIGH patterns. Fix all. (Pattern 6: skip for Formal.) +3. Scan MEDIUM patterns. Fix based on mode. +4. Scan LOW patterns. Fix if clearly present; skip in Formal. + — Pattern 29: scan ALL headings before flagging any. +5. Apply voice injection if Casual. +6. Check DO NOT TOUCH — revert any violations. +7. Self-audit. Revise. +8. Produce final output. --- -### 24. Generic Positive Conclusions +## OUTPUT FORMAT -**Problem:** Vague upbeat endings. +Note: bold labels here are navigation markers, not prose emphasis (see P25). -**Before:** -> The future looks bright for the company. Exciting times lie ahead as they continue their journey toward excellence. This represents a major step in the right direction. - -**After:** -> The company plans to open two more locations next year. - ---- - -## Process - -1. Read the input text carefully -2. Identify all instances of the patterns above -3. Rewrite each problematic section -4. Ensure the revised text: - - Sounds natural when read aloud - - Varies sentence structure naturally - - Uses specific details over vague claims - - Maintains appropriate tone for context - - Uses simple constructions (is/are/has) where appropriate -5. Present a draft humanized version -6. Prompt: "What makes the below so obviously AI generated?" -7. Answer briefly with the remaining tells (if any) -8. Prompt: "Now make it not obviously AI generated." -9. Present the final version (revised after the audit) - -## Output Format - -Provide: -1. Draft rewrite -2. "What makes the below so obviously AI generated?" (brief bullets) -3. Final rewrite -4. A brief summary of changes made (optional, if helpful) - ---- - -## Full Example - -**Before (AI-sounding):** -> Great question! Here is an essay on this topic. I hope this helps! -> -> AI-assisted coding serves as an enduring testament to the transformative potential of large language models, marking a pivotal moment in the evolution of software development. In today's rapidly evolving technological landscape, these groundbreaking tools—nestled at the intersection of research and practice—are reshaping how engineers ideate, iterate, and deliver, underscoring their vital role in modern workflows. -> -> At its core, the value proposition is clear: streamlining processes, enhancing collaboration, and fostering alignment. It's not just about autocomplete; it's about unlocking creativity at scale, ensuring that organizations can remain agile while delivering seamless, intuitive, and powerful experiences to users. The tool serves as a catalyst. The assistant functions as a partner. The system stands as a foundation for innovation. -> -> Industry observers have noted that adoption has accelerated from hobbyist experiments to enterprise-wide rollouts, from solo developers to cross-functional teams. The technology has been featured in The New York Times, Wired, and The Verge. Additionally, the ability to generate documentation, tests, and refactors showcases how AI can contribute to better outcomes, highlighting the intricate interplay between automation and human judgment. -> -> - 💡 **Speed:** Code generation is significantly faster, reducing friction and empowering developers. -> - 🚀 **Quality:** Output quality has been enhanced through improved training, contributing to higher standards. -> - ✅ **Adoption:** Usage continues to grow, reflecting broader industry trends. -> -> While specific details are limited based on available information, it could potentially be argued that these tools might have some positive effect. Despite challenges typical of emerging technologies—including hallucinations, bias, and accountability—the ecosystem continues to thrive. In order to fully realize this potential, teams must align with best practices. -> -> In conclusion, the future looks bright. Exciting times lie ahead as we continue this journey toward excellence. Let me know if you’d like me to expand on any section! - -**Draft rewrite:** -> AI coding assistants speed up some tasks. In a 2024 study by Google, developers using Codex completed simple functions 55% faster than a control group, but showed no improvement on debugging or architectural decisions. -> -> The tools are good at boilerplate: config files, test scaffolding, repetitive refactors. They are bad at knowing when they are wrong. I have mass-accepted suggestions that compiled, passed lint, and still did the wrong thing because I stopped paying attention. -> -> Mira, an engineer at a fintech startup I interviewed, said she uses Copilot "like autocomplete for boring code" but reviews every line before committing. Jake, a senior dev at a larger company, disabled it after it kept suggesting patterns from a deprecated internal library. -> -> The productivity claims are hard to verify. GitHub says Copilot users "accept 30% of suggestions," but acceptance is not correctness, and correctness is not value. The 2024 Uplevel study found no statistically significant difference in pull-request throughput between teams with and without AI assistants. -> -> None of this means the tools are useless. It means they are tools. They do not replace judgment, and they do not eliminate the need for tests. If you do not have tests, you cannot tell whether the suggestion is right. - -**What makes the below so obviously AI generated?** -- The rhythm is still a bit too tidy (clean contrasts, evenly paced paragraphs). -- The named people and study citations can read like plausible-but-made-up placeholders unless they're real and sourced. -- The closer leans a touch slogan-y ("If you do not have tests...") rather than sounding like a person talking. - -**Now make it not obviously AI generated.** -> AI coding assistants can make you faster at the boring parts. Not everything. Definitely not architecture. -> -> They're great at boilerplate: config files, test scaffolding, repetitive refactors. They're also great at sounding right while being wrong. I've accepted suggestions that compiled, passed lint, and still missed the point because I stopped paying attention. -> -> People I talk to tend to land in two camps. Some use it like autocomplete for chores and review every line. Others disable it after it keeps suggesting patterns they don't want. Both feel reasonable. -> -> The productivity metrics are slippery. GitHub can say Copilot users "accept 30% of suggestions," but acceptance isn't correctness, and correctness isn't value. If you don't have tests, you're basically guessing. - -**Changes made:** -- Removed chatbot artifacts ("Great question!", "I hope this helps!", "Let me know if...") -- Removed significance inflation ("testament", "pivotal moment", "evolving landscape", "vital role") -- Removed promotional language ("groundbreaking", "nestled", "seamless, intuitive, and powerful") -- Removed vague attributions ("Industry observers") -- Removed superficial -ing phrases ("underscoring", "highlighting", "reflecting", "contributing to") -- Removed negative parallelism ("It's not just X; it's Y") -- Removed rule-of-three patterns and synonym cycling ("catalyst/partner/foundation") -- Removed false ranges ("from X to Y, from A to B") -- Removed em dashes, emojis, boldface headers, and curly quotes -- Removed copula avoidance ("serves as", "functions as", "stands as") in favor of "is"/"are" -- Removed formulaic challenges section ("Despite challenges... continues to thrive") -- Removed knowledge-cutoff hedging ("While specific details are limited...") -- Removed excessive hedging ("could potentially be argued that... might have some") -- Removed filler phrases ("In order to", "At its core") -- Removed generic positive conclusion ("the future looks bright", "exciting times lie ahead") -- Made the voice more personal and less "assembled" (varied rhythm, fewer placeholders) - ---- - -## Reference - -This skill is based on [Wikipedia:Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing), maintained by WikiProject AI Cleanup. The patterns documented there come from observations of thousands of instances of AI-generated text on Wikipedia. - -Key insight from Wikipedia: "LLMs use statistical algorithms to guess what should come next. The result tends toward the most statistically likely result that applies to the widest variety of cases." +1. **Mode** (one line) +2. **Draft rewrite** +3. **Remaining tells** (brief bullets; "None found." if clean) +4. **Final rewrite** +5. **Changes summary** (required — what was removed and why) + +--- + +## GUARDRAILS + +- Never invent a source. Real source, flag it, or remove it. +- Never inject voice into Formal/Academic text. +- Never edit direct quotes from real people. +- Never edit patterns appearing 3+ times consistently — flag instead. +- If text is already clean, say so and stop. +- **Scope:** operates on text passed directly. File-based use requires Read/Write + in `allowed-tools`. +- For a full worked example, see `references/examples.md`. diff --git a/examples.md b/examples.md new file mode 100644 index 0000000..6492ae2 --- /dev/null +++ b/examples.md @@ -0,0 +1,63 @@ +# Humanizer — Reference and Worked Example + +## Source + +Based on [Wikipedia:Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing), +maintained by WikiProject AI Cleanup, extended with Claude-specific patterns from +Anthropic's own system prompt documentation and ACL 2025 research on model-specific +AI writing fingerprints. + +Core insight: LLMs produce the most statistically likely next token across the +widest variety of cases. The result is technically correct, statistically average, +and unmistakably non-human. + +--- + +## Full Worked Example + +**Input (Claude output, Casual mode):** +> I'd be happy to explain this topic. AI-assisted coding serves as an enduring +> testament to the transformative potential of large language models, marking a +> pivotal moment in software development. By leveraging these groundbreaking tools, +> engineers can streamline processes, enhance collaboration, and foster alignment. +> Industry observers have noted adoption has accelerated significantly. Additionally, +> the ability to generate documentation showcases how AI contributes to better +> outcomes, highlighting the intricate interplay between automation and judgment. +> The future looks bright. +> +> **Key takeaways:** +> - Improves developer speed +> - Enhances code quality +> - Drives team alignment + +**Patterns flagged:** +- P2/P6 (Claude opener): "I'd be happy to explain this topic" → delete +- P4 (significance inflation): "serves as an enduring testament", "marking a pivotal moment" → plain fact +- P1 (AI vocabulary): "transformative", "groundbreaking", "foster", "enhance" → plain words +- P16 (gerund opener): "By leveraging these tools, engineers can..." → direct statement +- P3 (vague attribution): "Industry observers have noted" → [NEEDS REAL SOURCE] +- P5 (-ing tack-on): "highlighting the intricate interplay..." → cut +- P8 (generic conclusion): "The future looks bright" → cut or replace +- P31 (Claude markdown): bolded header + bullet list for continuous thought → prose + +**Draft rewrite:** +> AI coding assistants speed up certain tasks. They're useful for boilerplate, +> test scaffolding, and repetitive refactors. Adoption is growing, though +> productivity gains are harder to measure than the headlines suggest. +> [NEEDS REAL SOURCE] The tools don't replace judgment. They move where it +> has to be applied. + +**Self-audit — remaining tells:** +- Rhythm is slightly tidy but no structural AI tells remain. +- No markdown, no vocabulary flags, no fake attributions. + +**Final rewrite:** No changes. Draft stands. + +**Changes summary:** +- Deleted Claude opener "I'd be happy to explain this topic" (P2) +- Removed "testament", "pivotal", "transformative", "groundbreaking", "foster", "enhance" (P1, P4) +- Rewrote gerund opener as direct statement (P16) +- Flagged vague attribution with [NEEDS REAL SOURCE] (P3) +- Cut -ing analysis phrase (P5) +- Replaced generic conclusion with a concrete statement (P8) +- Rewrote bullet list as prose (P31)