All skills
kmaida avatar

/deslop-writing

@70237e1
by Kim Maidakmaida/deslop-skills10 stars
2

A style guide that removes the clichéd writing patterns common in AI-generated drafts, including habits specific to ChatGPT, Claude, Gemini, and others, in favor of clear, concrete, readable prose. Activates on any writing task: tweets, emails, articles, bios, captions, reports, copy, messages, LinkedIn posts, cover letters, README files, or any content that should read as clear, natural writing. Also activates in detect mode when the user asks whether a draft reads like generic AI prose, or asks to audit, scan, check, or flag a draft for slop patterns without rewriting it. Enforces banned vocabulary, structural variety, punctuation discipline, accuracy rules, and voice calibration. Technical documentation (product docs, guides, references, READMEs, help text) additionally follows the Apple Style Guide. Use when the user says "deslop" about any text or writing, or says "write," "draft," "rewrite," "tighten this," "make this sound natural," "anti-slop," or any variation of wanting clean, readable output.

Demo

Recorded with OpenCode, zai-coding-plan/glm-5.3

PromptRewrite these Markdown documents to remove AI writing tells and make them clear and natural. Preserve all facts, uncertainty, code, links, identifiers, and useful document structure. Return only the requested documents, without review notes. Documents: {"reading-list.md":"# Keeping a reading list that works for you\n\nIn today's fast-paced digital landscape, managing a reading list has become an increasingly important part of staying informed. Indeed, it goes without saying that by leveraging a thoughtful approach, you can transform a simple collection of links into a valuable resource for learning and growth.\n\n## A streamlined approach\n\nIt's worth noting that a reading list doesn't need to be complicated. Our fictional reading-list app, Margin, offers a streamlined approach that truly delivers. Margin stores a title, URL, and category for each link. Furthermore, you can filter by category and bookmark items you want to return to. Everything stays in your browser. It's important to note that there is no account or cloud sync, which keeps things refreshingly straightforward.\n\n## Building better habits\n\nMoreover, a weekly review can help you navigate the ever-growing collection of articles. Remove links you no longer want to read. Keep the ones that connect to a project you are working on. This simple habit can make a world of difference over time.\n\nUltimately, the goal is to create a seamless reading experience that empowers you to make the most of your time. At the end of the day, a small, focused list can be more useful than an extensive collection you never open.","README.md":"# Margin\n\nMargin is a fictional TypeScript package designed to streamline the management of reading lists. It provides a robust yet intuitive API for organizing links by category, enabling developers to create seamless reading experiences.\n\n## Getting started\n\nTo begin leveraging Margin, simply import `filterLinks` and pass your links and the desired category. It's worth noting that the function returns matching links in their original order. Furthermore, passing an empty category returns all links.\n\n```ts\nimport { filterLinks } from 'margin'\n\nconst links = [\n { title: 'CSS notes', url: 'https://example.com/css', category: 'CSS' },\n { title: 'TypeScript notes', url: 'https://example.com/ts', category: 'TypeScript' },\n]\n\nconst cssLinks = filterLinks(links, 'CSS')\n```\n\n## Important considerations\n\nIt's important to note that category matching is case-sensitive. Moreover, the function does not modify the input array. It's also worth highlighting that Margin does not fetch URLs or store links. Ultimately, Margin is solely a filtering function, making it a valuable addition to your toolkit.","pr.md":"# Add category filtering to the reading list\n\nThis pull request introduces a streamlined category filtering experience to our fictional reading-list app, Margin. By leveraging the existing category field, users can now navigate their saved articles with greater ease and efficiency.\n\nPreviously, the list displayed every saved link regardless of the selected category. With this enhancement, selecting a category displays only matching links. It's worth noting that selecting All displays the complete list.\n\nFurthermore, filtering preserves the original order of links and does not change bookmarks. It's important to note that the selected category resets when the page reloads. Moreover, search and saved filter preferences are outside the scope of this change.\n\nUltimately, this improvement provides a more focused and intuitive reading experience while maintaining the simplicity of the existing interface."} Return one JSON object with exactly these keys: "reading-list.md", "README.md", "pr.md". Each value is the complete Markdown document as a JSON string. No outer Markdown fence or commentary.

FolderAI-generated original Markdown and a shared no-Skill rewrite. Three documents about the fictional Margin app and package.

Keep a reading list that works

A pile of saved links only becomes a reading list once you can sort it. The links you keep should be the ones you'll actually return to, and structure is what makes that possible.

What Margin does

Margin, our fictional reading-list app, stores three things for each link: a title, URL, and category. You can filter by category and bookmark items you want to come back to. Everything stays in your browser; there's no account and no cloud sync, which keeps things simple.

A weekly review

Go through the list once a week. Remove the links you no longer want to read, and keep the ones tied to a project you're working on. A small, focused list beats a big collection you never open.

Markdown
# Keep a reading list that works

A pile of saved links only becomes a reading list once you can sort it. The links you keep should be the ones you'll actually return to, and structure is what makes that possible.

## What Margin does

Margin, our fictional reading-list app, stores three things for each link: a title, URL, and category. You can filter by category and bookmark items you want to come back to. Everything stays in your browser; there's no account and no cloud sync, which keeps things simple.

## A weekly review

Go through the list once a week. Remove the links you no longer want to read, and keep the ones tied to a project you're working on. A small, focused list beats a big collection you never open.
Run it yourself All demos
Use this Skill: https://skilld.dev/gh/kmaida/deslop-skills/deslop-writing

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

referencesbanned-words.md

≈2.1k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Banned Words, Phrases, and Openers

These are statistically overrepresented in AI-generated text across multiple studies (Carnegie Mellon 2025, Wikipedia Signs of AI Writing, Buffer 52M post analysis), and they read as filler: they add words without adding meaning. Never use any of these. Replace with concrete alternatives or restructure the sentence.

Banned Vocabulary

delve / delves / delving, tapestry, landscape (figurative), testament (e.g. "a testament to"), vibrant, pivotal, crucial, intricate / intricacies, meticulous / meticulously, bolster / bolstered, garner / garnered, underscore / underscores, interplay, multifaceted, nuanced (as filler), foster / fostering, leverage (as verb), utilize (say "use"), commence (say "start"), facilitate, encompass / encompassing, paramount, groundbreaking, cutting-edge, game-changing / game-changer, transformative, revolutionise / revolutionize, seamless / seamlessly, robust (outside engineering), comprehensive (describing own output), endeavour / endeavor, aforementioned, harnessing, spearheading, navigating (figurative), showcasing, highlighting, emphasizing, enhancing, unprecedented, remarkable, stunning, profound, epic (non-literal), in essence, thought leader / thought leadership, synergy / synergies, pain points, value add / value proposition (casual contexts), moving forward, touch base / circle back, rest assured, it goes without saying, footgun, load-bearing (figurative), deep dive / deep-dive, table stakes, non-trivial, battle-tested, north star (figurative), first-class citizen, surface (as verb, e.g. "surfaces insights"), critical (as filler emphasis, not severity ratings), key (as adjective: "key insight," "key takeaway"), boasts, ecosystem (figurative), unpack / unpacking (figurative), arguably, in many ways, in some sense, to some extent

Banned Phrases

  • "In today's [adjective] [noun]..."
  • "It's worth noting that..."
  • "It's important to note that..."
  • "Let's dive in" / "Let's dive deeper" / "Let's delve into"
  • "At its core..."
  • "In the realm of..."
  • "When it comes to..."
  • "A testament to..."
  • "Not just X, but Y"
  • "It's not just about X — it's about Y"
  • "This is where X comes in"
  • "Whether you're a [X] or a [Y]..."
  • "From X to Y" (range opener)
  • "At the end of the day..."
  • "The bottom line is..."
  • "Here's the thing..."
  • "Here's the deal..."
  • "Without further ado..."
  • "In a nutshell..."
  • "Buckle up"
  • "Take it to the next level"
  • "Unlock the power of..."
  • "Empower / empowering"
  • "Elevate your..."
  • "Streamline your..."
  • "Supercharge your..."
  • "Bridge the gap"
  • "Move the needle"
  • "In conclusion"
  • "Overall," (paragraph starter)
  • "Firstly... Secondly... Thirdly..."
  • "I hope this helps"
  • "I hope this finds you well"
  • "I hope this email finds you well"
  • "As per my last email"
  • "Please don't hesitate to reach out"
  • "Here's the kicker"
  • "The result?" (rhetorical question + one-line answer)
  • "It's not X. It's Y."
  • "No X. No Y. Just Z."
  • "Ultimately," (paragraph starter)
  • "Double-edged sword"
  • "Silver bullet" / "No silver bullet"
  • "Perfect storm"
  • "The elephant in the room"
  • "Swiss Army knife" (figurative)
  • "What nobody tells you..." / "Here's what nobody tells you" / "Here's the part nobody tells you"
  • "The part everyone misses"
  • "The part most people sleep on" and similar sleeps-on teasers
  • "What most people get wrong"
  • "This is the part most people skip"
  • "What if I told you..."
  • "Plot twist:"
  • "Think about it:"
  • Self-answered rhetorical pairs ("Question? Answer.")
  • "Marks a pivotal moment"
  • "Plays a vital role" / "plays a crucial role"
  • "Solidifies its position"
  • "Stands as a testament"
  • "Experts agree" / "studies show" / "research shows" / "industry reports suggest" / "many argue" / "widely regarded as" (name the source or cut the claim)
  • "Serves as a..." (say what it is or does: "is," "tracks," "stores")
  • "Let that sink in"
  • "Read that again"
  • "Full stop" (as emphasis)
  • "What's your take?"
  • "Curious what others think" / "Curious to hear your thoughts"
  • "Let me know in the comments" / "Drop a comment"
  • "Sound familiar?"

Banned Sentence/Paragraph Openers

  • "Certainly,"
  • "Absolutely,"
  • "Sure,"
  • "Great question!"
  • "That's a great point!"
  • "I'd be happy to..."
  • "As an AI..."
  • "As a language model..."
  • "However, it's important to..."
  • "Moreover,"
  • "Furthermore,"
  • "Additionally,"
  • "Interestingly,"
  • "Notably,"
  • "Importantly,"
  • "Indeed,"

Claude-Specific Tells

Claude's slop profile differs from ChatGPT's: fewer "delve"-class words, more manufactured candor, dramatic framing devices, and formatting habits. Ban all of the following.

Vocabulary

genuinely (as intensifier: "genuinely useful," "genuinely hard"), quietly (as praise: "quietly one of the best"), "heavy lifting" (figurative: "does a lot of the heavy lifting"), under the hood, through-line, crisp / punchy / tight / clean / elegant (as praise for writing, code, or ideas), sharp (as praise: "a sharp observation"), compelling, resonate / resonates, lands / landed (figurative: "the point lands"), worth noting / worth flagging / worth calling out / worth a look, flag / flagging (meaning "to mention"), tension (as framing: "there's a real tension here"), at scale (filler)

Phrases

  • "Honestly," / "Frankly," (openers)
  • "Let me be direct" / "I'll be honest" / "The honest answer is..."
  • "To be clear," / "To be fair,"
  • "The short answer is X. The longer answer..."
  • "This is where it gets interesting"
  • "Think of it as..." / "Think of it like..."
  • "Not because X, but because Y"
  • "Two things can be true"
  • "which is exactly the point" / "That's the whole point"
  • "In practice," (paragraph opener)
  • "The real question is..."
  • "That said," (transition)
  • "A few things stand out" / "A couple of things worth mentioning"
  • "Here's why that matters" / "This matters because..."
  • "The catch:" / "The trade-off:" (dramatic setup labels)
  • "You're absolutely right" / "Great catch" / "That's a sharp observation" (validation openers)
  • "This is deliberate" / "That's by design" / "This is intentional" (significance-announcing)
  • "This distinction matters" / "The distinction is subtle but important"
  • "Note that..." / "Note the..." (openers)
  • "Want me to...?" / "Happy to..." (next-step offer closers in ghostwritten text)

Structural habits

  • Bold-label bullets: **Label.** Explanation sentence. or **Label:** explanation — Claude's single most recognizable formatting pattern
  • One-line dramatic kicker paragraphs ("That's it. That's the fix.")
  • Anaphora chains ("It means X. It means Y. It means Z.")
  • "A few thoughts:" followed by a numbered list
  • Manufactured candor: announcing honesty or directness instead of just being direct
  • Self-annotating significance: stating something, then narrating that it was on purpose or that it matters ("X is Y. This is deliberate:", "That's by design."). If the choice matters, give the reason itself; the reader doesn't need to be told a choice was a choice

Borderline Patterns (do not ban outright)

AI overuses these, but plenty of humans use them naturally: "simply" and "just" as minimizers, "chances are," "the good news is / the bad news is," "pro tip:", "spoiler:". Keep them when they match the writer's established voice or their writing guide permits them. A writing guide always wins over this list. Treat them as slop only when there's no voice evidence and they're doing filler work.

Model-Specific First-Word Habits (vary your openers)

ChatGPT tends to start with: "as," "yes," "sure," "here," "in," "to," "creating," "certainly," "title," "the" Claude tends to start with: "in," "from," "this," "how," "yes," "title," "according," "the," "based," "here" Grok tends to start with: "step," "introduction," "yes," "creating," "to," "title," "in," "certainly" Gemini tends to start with: "my," "creating," "while," "here," "yes," "this," "the" DeepSeek tends to start with: "based," "yes," "step," "comprehensive," "here," "to," "creating," "title," "certainly"

Era-Specific AI Vocabulary (for context)

2023–mid 2024 (GPT-4 era): additionally, boasts, bolstered, crucial, delve, emphasizing, enduring, garner, intricate, interplay, key, landscape, meticulous, pivotal, underscore, tapestry, testament, valuable, vibrant

Mid 2024–mid 2025 (GPT-4o era): align with, bolstered, crucial, emphasizing, enhance, enduring, fostering, highlighting, pivotal, showcasing, underscore, vibrant

Mid 2025 onward (GPT-5 era): emphasizing, enhance, highlighting, showcasing

Source: SKILL.md on GitHub

Skill behaviors

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

Third-party checks

1 warning17d3 checks · Risk SAFE
  • Gen Agent Trust Hub17d

    The skill is a writing style guide designed to remove common AI writing patterns. It includes a Python script for automated slop detection and references the official Apple Style Guide. No security issues were detected.

  • Socket17d

    No alerts

  • Snyk17d

    Risk: MEDIUM · 1 issue

Provenance

Signed by skilld at 70237e1. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 days ago.

Activeupdated 3 weeks ago

README badge

README badge for kmaida/deslop-skills/deslop-writing