Strip 85 stylometric, syntactic, cross-linguistic, and marketing-hype AI writing tells across any language. Sub-skill: load only when routed from the hook-everything entry skill, not for general writing or editing.
Installs into .claude/skills of the current project.
Are you the author of Anti Slop?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/vumb-vibecoder-anti-slop)
---
name: anti-slop
description: Strip 85 stylometric, syntactic, cross-linguistic, and marketing-hype AI writing tells across any language. Sub-skill: load only when routed from the hook-everything entry skill, not for general writing or editing.
---
# Universal Anti-Slop & Anti-Hype (85 Deep Stylometric & Cognitive Rules)
Why can experienced readers still "smell" AI writing or cheap marketing even after you delete obvious words like `delve`, `tapestry`, and `seamless`? Because Large Language Models have a deeper **cognitive and syntactic signature**: uniform information entropy per sentence, see-saw concession clauses, polite professorial neutrality, abstract noun chains, symmetrical paragraph geometry, and forced trend-jacking.
Apply all 85 rules across **any human language** (English, Vietnamese, Spanish, French, German, Japanese, Chinese, Korean) to strip both surface cliches and deep structural AI tells.
---
## Dimension 1: Information Entropy & Discourse Geometry (Rules 1-10)
1. **Kill Uniform Information Entropy** because LLMs distribute information evenly so every sentence carries roughly the same lukewarm weight. Human writers compress high-surprise facts into dense bursts, followed by ultra-short punchlines.
2. **Ban the 3-Sentence "Sandwich Paragraph"** where LLMs follow the exact formula `[Sentence 1: Abstract claim] -> [Sentence 2: Dual/triple example] -> [Sentence 3: Moralizing summary tail]`. Break this mold: let a paragraph be a single 6-word verdict, or a 5-sentence causal narrative.
3. **Vary Paragraph Visual Density** by never writing four consecutive paragraphs of 3-4 lines each. Alternate short and long blocks so the page looks human-typeset before a single word is read.
4. **Delete Throat-Clearing Macro Openers** that announce the world is changing or a topic is important (`"In today's..."`, `"Trong bối cảnh..."`, `"En el mundo actual..."`). Start with a concrete actor, number, or event in Sentence 1.
5. **Delete Summary Echo Closers** starting with `"In conclusion,"`, `"Ultimately,"`, `"All in all,"`, or `"Tóm lại,"` that repeat the thesis. End at peak utility or on a concrete image.
6. **Eliminate Heading-to-First-Sentence Echo** by never repeating the words of an H2/H3 heading in the first sentence below it. Jump straight into the mechanism or story.
7. **Ban Colon-Split Brochure Titles** separated by a colon (`"The Future of Caching: How Redis Transforms Latency"`, `"Tối ưu hóa dữ liệu: Chìa khóa cho doanh nghiệp"`). Write single-clause declarative or tension headings (`"Why Redis evicted our session keys at 60% RAM"`).
8. **Break the Rule-of-Three Reflex** by restricting symmetrical triplets (`A, B, and C`) to at most once per 500 words. Use a single exact word, a 2-item contrast, or an uneven 4-item list.
9. **Ban Bold-Lead Bullet Soup** that formats explanations as vertical lists of `- **Bold Label**: one-sentence generic definition`. Write connected prose paragraphs for causal explanations and use compact tables for comparisons.
10. **Kill Formulaic "Challenges and Future Outlook" Tail Sections** that append unrequested closing sections titled `"Challenges and Limitations"` followed by `"Future Prospects"`. Embed trade-offs directly inside the main workflow steps.
---
## Dimension 2: Deep Syntactic & Clause Architecture (Rules 11-22)
11. **Ban Negative Parallelisms ("Not Just X, It's Y")** across all languages by cutting `"It's not just X, it's Y"`, `"More than merely X, it is Y"`, `"Không chỉ X mà còn Y"`, and `"No es solo X, sino Y"`. State directly what the thing does.
12. **Kill See-Saw Concession Openers ("While X, Y")** that start every third sentence with a balanced concession clause such as `"While X offers significant benefits, Y remains..."`, `"Although A is simple, B provides..."`, or `"Mặc dù X rất quan trọng, nhưng Y..."` *(English gloss: "Although X is very important, Y...")*. State the boundary condition directly (`"X works up to 10k rows. Past 10k rows, use Y."`).
13. **Chop Superficial Participial / Gerund Commentary Tails** attached after a comma to explain the "meaning" of the sentence (`"..., highlighting the need for..."`, `"..., ensuring seamless..."`, `"..., reflecting a broader shift..."`, `"..., qua đó khẳng định..."`, `"..., góp phần nâng cao..."`).
14. **Stop Copula Inflation (Use Plain "Is" and "Has")** by replacing `"serves as"`, `"stands as"`, `"functions as"`, `"boasts a"`, `"represents a"`, `"đóng vai trò như một"`, and `"được xem như là"` with plain `is`, `are`, `has`, `là`, or `có`.
15. **Ban Rhetorical Self-Q&A Micro-Clichés** where LLMs fake conversational punch by asking 2-word questions (`"The result? Faster builds."`, `"The catch? It costs more."`, `"The best part?"`, `"Why does this matter?"`, `"Here's the kicker:"`, `"Kết quả là gì?"`). Write normal, connected clauses instead.
16. **Kill Nominalization Chains (Abstract Noun Subjects)** by replacing `"The utilization of batch processing facilitates the minimization of network overhead"` or `"Việc áp dụng quy trình mới mang lại sự cải thiện cho hiệu suất"` *(English gloss: "The application of the new process brings an improvement to performance")* with active human syntax: `"Batching requests cuts network overhead in half"` / `"Gộp request giúp giảm một nửa băng thông"` *(English gloss: "Batching requests cuts bandwidth in half")*.
17. **Eliminate Passive Agency Erasure** such as `"Mistakes were made during the migration"` or `"A decision was reached to deprecate..."`. Name the actor (`"We dropped the wrong table"`, `"The core team deprecated..."`).
18. **Prevent Translationese & Pronoun Stuffing in Pro-Drop Languages** when writing in Vietnamese, Japanese, Korean, or Spanish. Never mechanically translate English passive voice (`"được thiết kế để được sử dụng bởi..."`) or stuff pronouns (`"bạn"`, `"chúng tôi"`, `"nó"`, `"của bạn"`) into every clause where native speakers naturally omit them.
19. **Ban False Universal Ranges ("From X to Y")** such as `"from solo developers to Fortune 500 enterprises"`, `"from ideation to execution"`, or `"từ người mới bắt đầu đến chuyên gia"`. Name the exact target cohort.
20. **Strip Mechanical Transition Stacking** by deleting sentence-initial `"Furthermore,"`, `"Moreover,"`, `"Additionally,"`, `"Consequently,"`, `"In essence,"`, `"At its core,"`, `"Hơn thế nữa,"`, and `"Bên cạnh đó,"`. Let logic carry the transition.
21. **Avoid Synonym Cycling (Elegant Variation)** where a writer calls a tool `"the queue"` in line 1, `"this messaging backbone"` in line 2, and `"the asynchronous pipeline"` in line 3. Repeat `"the queue"` or `"it"`.
22. **Cut Preparatory "It" & Cleft-Sentence Padding** by replacing `"It is important to recognize that it is the network layer that causes..."` with `"The network layer causes..."`.
---
## Dimension 3: Stylometric Rhythm & Burstiness Math (Rules 23-30)
23. **Enforce High Sentence-Length Burstiness (`CV >= 0.35`, `SD >= 5.0` words)** by never writing five consecutive medium-length sentences (15-22 words). Pair a 28-word sentence that traces a multi-step mechanism with a 3-to-5 word sentence that lands the consequence.
24. **Vary Syntactic Tree Depth** by alternating right-branching sentences (subject + verb + object + detail) with occasional direct imperatives (`"Check the logs first."`) or short conditional rules (`"If p99 exceeds 200ms, drop the cache."`).
25. **Ban Consecutive Sentence-Opener Repetition** by never starting two consecutive sentences with the exact same word (`The... The...`, `This... This...`, `We... We...`) unless using deliberate anaphora in a speech.
26. **Strict Em-Dash Quota (`< 2` per 1,000 words)** because LLMs use em-dashes (`—` or ` -- `) to glue clauses together lazily. Use periods, parentheses, or colons instead.
27. **Ban Semicolon Chains in Non-Academic Prose** instead of joining balanced generic clauses; split them with a period.
28. **Use Natural Contractions** in English (`it's`, `don't`, `won't`, `we've`, `didn't`). Zero contractions in everyday prose sounds like a stiff translation.
29. **Enforce List-Item Length Asymmetry** when using bullet points so not every bullet is 18-22 words long. Let one item be 4 words and another be 32 words based on real substance.
30. **Eliminate Triad Adjective Stacks** by never placing three adjectives before a noun (`"a fast, scalable, and intuitive CLI"`). Use one verifiable qualifier (`"a 4MB single-binary CLI"`).
---
## Dimension 4: Epistemic & Pragmatic Voice - Killing the "Polite Professor" (Rules 31-42)
31. **Ban Significance Inflation** that announces something is `"crucial"`, `"pivotal"`, `"vital"`, `"a testament to"`, `"transformative"`, or `"then chốt"`. Show the concrete cost of ignoring it.
32. **Ban Vague Weasel Attributions** such as `"Experts argue"`, `"Studies show"`, `"Industry leaders agree"`, or `"Nhiều chuyên gia cho rằng"`. Name the author, paper, or dataset, or own the claim directly.
33. **Zero Chatbot Meta-Frames & Sycophancy** by stripping `"Certainly!"`, `"Great question!"`, `"Here is a breakdown:"`, `"Let's dive in"`, `"I hope this helps"`, and `"Without further ado"`.
34. **Eliminate Micro-Hedging Stacks** by cutting cowardly qualifiers like `"tends to somewhat"`, `"can often serve to"`, and `"may potentially help"`. State the exact condition under which the statement holds true.
35. **Kill Fence-Sitting Both-Sidesism** that concludes a comparison with `"Ultimately, the right choice depends on your unique business needs."` Give a sharp decision rule: `"Pick A if your team is under 5 people; pick B once write volume passes 1,000 req/sec."`
36. **Replace Unquantified Intensifiers with Numbers** by banning `"blazingly fast"`, `"massively scalable"`, `"dramatically reduced"`, and `"vô cùng hiệu quả"`. State the before-and-after numbers.
37. **Ban Self-Referential Artifact Narration** such as `"This article explores..."`, `"In this section, we will discuss..."`, or `"This commit adds..."`. Start with the subject itself.
38. **Kill Toxic Positivity / Emotional Sanitization** where LLMs turn every failure into a cheerful moral lesson (`"While the outage was challenging, it provided an invaluable learning opportunity for our team"`). Humans admit when something simply sucked (`"The outage cost us $18,000 and ruined our weekend. We changed two rules on Monday so it wouldn't happen again."`).
39. **Ban Pedagogical Lecture Transitions** like `"To truly understand X, one must first examine Y"` or `"Before diving into X, let's take a step back and define Y"`. Assume the reader is a smart peer.
40. **Ban Pompous Coined Frameworks** that invent capitalized buzzword labels for ordinary actions (e.g., calling "checking your logs" `"The Proactive Observability Paradigm"`).
41. **Avoid Moralizing "Should / Must" Preaching** by showing the consequence (`"Leave port 6379 open without a password and a crypto-miner will find your Redis instance in under 12 minutes"`) instead of lecturing (`"Developers must always prioritize security"`).
42. **Preserve Verbatim Quotes Untouched** by never paraphrasing or "cleaning up" text inside quotation marks or blockquotes.
---
## Dimension 5: Abstract-to-Concrete Ratio (ACR) & Sensory Grounding (Rules 43-50)
43. **The 35-Word Concrete Floor** requires every 35 words of prose to contain at least one concrete anchor (a physical noun, an exact number/unit, a specific file/tool name, or a visible human action). If a 40-word span contains only abstract nouns (`strategy, alignment, optimization, ecosystem, value`), rewrite it immediately.
44. **Ban "Ghost Actors"** so `the landscape`, `the ecosystem`, `modern organizations`, or `stakeholders` never serve as the subject of a sentence. Name the actual person (`the on-call engineer`, `your CFO`, `a first-time buyer`).
45. **Replace Cliché Metaphors with Mechanical Analogies** by banning `tapestry`, `beacon`, `north star`, `double-edged sword`, `tip of the iceberg`, `chìa khóa vàng`, and `kim chỉ nam`. If you use an analogy, make it mechanical (`"acts like a check valve that only lets traffic flow one way"`).
46. **Ground Time References** by replacing `"recently"`, `"in the past"`, and `"nowadays"` with exact temporal anchors (`"in Q3"`, `"after Postgres 16 shipped"`, `"at 2:14 AM Tuesday"`).
47. **Ground Failure Modes** by replacing `"performance may degrade"` with `"latency spikes from 40ms to 4.2s and the tab freezes"`.
48. **Ground Human Behavior in Muscle Memory** by describing the physical action (`"saves you from scrolling horizontally across 19 columns to find the Delete button"`) instead of `"improves user experience"`.
49. **Ban Vaporware Verbs** including `empower`, `unlock`, `unleash`, `elevate`, `foster`, `spearhead`, `harness`, `revolutionize`, `catalyze`, and `streamline`. Use physical or operational verbs (`cut`, `block`, `merge`, `strip`, `route`, `cache`, `split`).
50. **Ban Empty Determiner Phrases** such as `"a wide range of"`, `"a variety of"`, `"numerous different"`, and `"nhiều loại ... khác nhau"`. Name 2 specific examples or give the exact count.
---
## Dimension 6: Multi-Language Lexical & Calque Blacklists (Rules 51-60)
51. **English Tier-1 Instant-Ban Words** include `delve`, `tapestry`, `realm`, `pivotal`, `seamless`, `seamlessly`, `robust`, `game-changer`, `leverage` (as verb), `elevate`, `unlock`, `unleash`, `landscape` (figurative), `crucial`, `meticulous`, `multifaceted`, `holistic`, `synergy`, `paradigm`, `beacon`, `cornerstone`, `catalyst`, `plethora`, `myriad`, and `indelible`.
52. **English Tier-2 Phrase Bans** include `"in today's fast-paced"`, `"it's important to note"`, `"plays a vital/pivotal/crucial role"`, `"stands/serves as a testament"`, `"sits at the intersection of"`, `"strikes a balance"`, `"at the heart of"`, `"paves the way for"`, `"a deep dive into"`, and `"in the wake of"`.
53. **Vietnamese Tier-1 "AI Boilerplate" Bans** include `"trong bối cảnh"` *("in the context of")*, `"trong kỷ nguyên số"` *("in the digital era")*, `"không thể phủ nhận"` *("it is undeniable")*, `"như chúng ta đã biết"` *("as we all know")*, `"hãy cùng khám phá"` *("let's explore")*, `"đóng vai trò quan trọng/then chốt"` *("plays a vital/pivotal role")*, `"mang đến trải nghiệm mượt mà/toàn diện"` *("delivers a seamless/holistic experience")*, `"chìa khóa vàng"` *("golden key")*, `"kim chỉ nam"` *("guiding compass / north star")*, `"mở ra cánh cửa"` *("opens the door to")*, `"khẳng định vị thế"` *("affirms the position")*, `"tạo tiền đề"` *("creates the premise / paves the way")*, `"góp phần nâng cao"` *("contributes to enhancing")*, `"hành trình khám phá"` *("journey of discovery")*, and `"bức tranh toàn cảnh"` *("the big picture / panorama")*.
54. **Vietnamese Tier-2 "Machine-Translated Boilerplate" Bans** include `"Không chỉ ... mà còn ..."` *("Not only ... but also ...")*, `"Một sự [Verb] đã được..."` *("A [nominalized verb] was performed...")*, `"Việc thực hiện [Verb]..."` *("The implementation of [Verb]...")*, `"Tóm lại, có thể thấy rằng..."` *("In conclusion, it can be seen that...")*, plus unnatural repetition of pronouns `"bạn / của bạn / chúng tôi"` *("you / your / we")* in every clause where native Vietnamese omits the pronoun.
55. **Spanish AI-Slop Bans** include `"en el vertiginoso mundo actual"`, `"juega un papel fundamental/crucial"`, `"no es solo X, sino Y"`, `"un tapiz de"`, `"es importante destacar que"`, and `"en conclusión"`.
56. **French AI-Slop Bans** include `"dans un monde en constante évolution"`, `"joue un rôle crucial/clé"`, `"ce n'est pas seulement X, c'est Y"`, `"il est important de noter que"`, and `"une pléthore de"`.
57. **German AI-Slop Bans** include `"in der heutigen schnelllebigen Welt"`, `"spielt eine entscheidende Rolle"`, `"nicht nur X, sondern auch Y"`, and `"es ist wichtig zu beachten"`.
58. **CJK (Chinese / Japanese / Korean) AI-Slop Bans** cover Chinese (`在当今快节奏的`, `不仅仅...而是...`, `发挥着至关重要的作用`, `总而言之`, `让我们一起探索`), Japanese (`〜において重要な役割を果たしています`, `単なる〜ではなく`, `〜と言っても過言ではありません`, `結論として`), and Korean (`오늘날 빠르게 변화하는`, `단순히 ~일 뿐만 아니라`, `중요한 역할을 합니다`).
59. **Code & Commit Message AI Tells** in technical repos ban `"This commit leverages/enhances..."`, `"comprehensive error handling"`, `"production-ready"`, `"blazingly fast"`, and trailing `", ensuring consistency"` in commit messages.
60. **Emoji & Typographic Clutter** bans rocket ships, sparkles, fire, lightbulbs, and green checkmarks in prose, headings, and professional copy.
---
## Dimension 7: Positive Human Texture & Cognitive Authenticity (Rules 61-70)
To make writing indistinguishable from a sharp human practitioner, inject at least 3 of these 10 markers:
61. **The Mid-Thought Self-Correction** qualifies a thought the way a real person thinks in real time (`"That sounds like a network bug - and for the first two hours we thought it was - until we checked the disk inode table."`).
62. **Candid Parenthetical Aside** drops a dry, realistic detail in parentheses `(mostly because nobody wanted to touch the 2019 billing cron job on a Friday afternoon)`.
63. **Earned Understatement Over Hyperbole** understates the adjectives when a result is impressive and lets the raw numbers do the work (`"Cutting the payload from 1.8MB to 14KB made the mobile app noticeably less annoying to open on train Wi-Fi."`).
64. **Specific Annoyance / Friction Detail** names the small, lived friction that proves you actually did the work (`"having to paste a 64-character token into three different .env files"`).
65. **First-Principles Causal Chain** explains *how* A causes B mechanically (`"Because Node uses a single event loop, parsing a 40MB JSON string blocks every incoming health-check ping for 600ms"`).
66. **The 1-to-4 Word Verdict Sentence** lands a complex paragraph with a blunt short sentence (`"It didn't work."`, `"That fixed it."`, `"Dead simple."`).
67. **Direct Second-Person Muscle-Memory Callout** references an exact micro-experience the reader recognizes (`"You know that sinking feeling when git rebase stops on file 14 of 40..."`).
68. **Dry Contrast Humor** pairs a grand expectation with a mundane reality (`"We architected the queue for 100,000 events per second. Our first paying customer sent 12 events a day, all at 9:00 AM, and still managed to break our parser with a semicolon in their company name."`).
69. **Un-Rounded Forensic Numbers** prefer real, un-rounded measurements (`14.2%`, `38 minutes`, `19 tables`) over suspiciously round numbers (`10%`, `1 hour`, `20 tables`).
70. **Abrupt High-Signal Closer** stops writing the exact moment the final useful point lands, without adding a "warm sign-off" paragraph.
---
## Dimension 8: Execution & Anti-Drift Protocol (Rules 71-75)
71. **The 300-Word Block Audit (Anti-Drift Rule)** prevents LLM stylistic decay after paragraph 4 on long texts (>400 words) by auditing and rewriting in 300-word chunks so paragraph 10 is just as sharp as paragraph 1.
72. **Restructure Clauses, Never Just Swap Synonyms** because swapping `"leverage"` for `"use"` inside a robotic sentence leaves the AI sentence skeleton intact. When a sentence triggers `slop_scan.py`, rewrite the subject and verb from scratch.
73. **The First-and-Last Sentence Compression Test** inspects the first sentence and the last sentence of the draft; if deleting either one loses zero factual information, delete them both.
74. **The Coffee-Shop Read-Aloud Test** reads every line aloud; if you would feel embarrassed saying a phrase to a respected colleague across a coffee table, cut it.
75. **Mandatory Automated Verification (`tools/slop_scan.py`)** runs `python3 tools/slop_scan.py <file> --fix-hints` whenever shell execution is available, revising until the score is `<= 15/100` (`CLEAN`).
---
## Dimension 9: Forced Trend-Jacking & Scam/Hype Marketing Tells (Rules 76-85)
76. **Ban Forced News-Peg Openers** such as `"With the recent rise/explosion of [Trend]..."`, `"As we saw in the news this week..."`, `"In light of recent events..."`, `"Trước làn sóng..."` *("Facing the wave of...")*, and `"Với sự bùng nổ của..."` *("With the explosion of...")*. If a market shift matters (`skills/market-intelligence`), state its direct operational effect on the user's workflow without announcing the trend.
77. **Ban Unrelated Metaphor Trend-Jacking** that compares a mundane tool or topic to a trending celebrity, viral meme, or macro headline without a direct causal link (Organic Fit Gate).
78. **Ban Scam ROI Multipliers** such as `"10x your productivity/revenue"`, `"guaranteed ROI"`, `"skyrocket your sales"`, `"explode your growth"`, `"x10 doanh thu"` *("10x revenue")*, and `"cam kết lợi nhuận"` *("guaranteed profit")*.
79. **Ban "Secret Blueprint" Guru Framing** including `"the secret blueprint"`, `"what the gurus won't tell you"`, `"the underground method"`, `"bí mật triệu đô"` *("million-dollar secret")*, and `"công thức bí truyền"` *("secret formula")*.
80. **Ban Manufactured Scarcity Phrases** such as `"spots are filling up fast"`, `"act now before it's too late"`, `"once-in-a-lifetime opportunity"`, `"chỉ còn duy nhất X suất"` *("only X spots left")*, and `"đăng ký ngay kẻo lỡ"` *("register now before you miss out")*.
81. **Ban Manipulative Decline Buttons (Confirmshaming)** like `"No thanks, I prefer wasting time"`.
82. **Ban Fake Authority Name-Dropping** such as `"Used by top engineers at Google, Meta, and Netflix"` unless citing a verifiable enterprise agreement or public repo usage.
83. **Ban Income / Lifestyle Bait** including `"financial freedom"`, `"passive income machine"`, `"work from the beach"`, and `"tự do tài chính"` *("financial freedom")*.
84. **Ban Exclamation-Mark Inflation** by keeping 0 exclamation marks in B2B/technical copy and at most 1 per 1,000 words in consumer copy. Never end a headline with `!`.
85. **Ban All-Caps Urgency Prefixes** such as `"URGENT:"`, `"BREAKING:"`, `"WARNING:"`, or `"LAST CHANCE:"`.