{"slug": "one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to", "title": "One Open Source Project a Day (No. 78): stop-slop - A Skill File That Teaches AI to Eliminate Its Own Writing Tells", "summary": "Product designer Hardik Pandya created stop-slop, an open-source Skill file that teaches AI writing tools to recognize and eliminate their own predictable writing patterns. The project, which has garnered 5,800 stars on GitHub, contains eight rules and a scoring rubric designed to strip out common AI tells such as throat-clearing openers, binary contrasts, and dramatic standalone sentences.", "body_md": "\"If it sounds like a pull-quote, rewrite it.\"\n\nThis is the 78th article in the \"One Open Source Project a Day\" series. Today's project is **stop-slop**.\n\nYou know the writing. It opens with \"In today's fast-paced world...\" The second paragraph has \"it's worth noting that.\" The middle is full of \"undeniably,\" \"transformative impact,\" \"revolutionary breakthrough.\" It ends with a profound one-liner. You recognize it instantly — this was written by an AI.\n\nThese patterns have become the signature tells of AI-generated content. **stop-slop** is a Claude Code Skill file with a simple mission: **teach AI to recognize and actively eliminate its own writing tells**. 5.8k Stars, pure Markdown, no code, works in any AI tool that supports custom system prompts.\n\nstop-slop was created by product designer Hardik Pandya. He's a serious writer who, after deep use of AI-assisted writing, noticed a problem: **AI-generated text follows a highly predictable set of patterns** — regardless of topic, regardless of prompt, the same opening styles appear, the same pivot structures, the same \"grand narrative\" tone.\n\nHis solution wasn't \"make AI write better in some abstract sense.\" It was to **explicitly enumerate all the known AI writing tells, then tell the AI: when you see these, fix them.**\n\nThe execution is rigorous — 8 rules, 3 reference files, a pre-publish checklist, a scoring rubric — forming a complete system that can be embedded into any writing workflow.\n\nstop-slop is a Skill file. Once installed, it activates automatically during writing or editing tasks and systematically identifies and removes AI writing tells.\n\n```\nYour prose draft\n      ↓\nstop-slop Skill (activates)\n      ↓\n  Identifies 8 categories of AI writing patterns\n      ↓\n  Corrects each one\n      ↓\n  5-dimension score (< 35/50 = needs revision)\n      ↓\nOutput: writing that sounds like a human wrote it\n```\n\n**Claude Code (Skill directory install)**:\n\n```\n# Clone into Claude's global skills directory\ngit clone https://github.com/hardikpandya/stop-slop.git ~/.claude/skills/stop-slop\n\n# After installation, the Skill activates automatically during writing tasks\n```\n\n**Claude Projects / Custom Instructions**:\n\n```\n# Copy the contents of SKILL.md into \"Custom Instructions\" or \"System Prompt\"\n```\n\n**Direct API**:\n\n``` python\nimport anthropic\n\nwith open(\"SKILL.md\", \"r\") as f:\n    skill_content = f.read()\n\nclient = anthropic.Anthropic()\nresponse = client.messages.create(\n    model=\"claude-sonnet-4-6\",\n    system=skill_content,\n    messages=[{\"role\": \"user\", \"content\": \"Please edit this passage: ...\"}]\n)\n```\n\nThe heart of SKILL.md. Each rule targets a specific category of AI writing patterns:\n\n**Rule 1: Cut Filler**\n\n```\nTargets: throat-clearing openers, emphasis crutches, all adverbs\n\nTypical AI tells:\n  ✗ \"In today's rapidly evolving digital landscape...\"\n  ✗ \"It's worth noting that\"\n  ✗ \"Certainly,\" \"Indeed,\" \"Obviously,\" \"Very,\" \"Extremely\"\n  ✗ \"It goes without saying\"\n\nFix: Start with the actual point. Remove all adverbs.\n```\n\n**Rule 2: Break Formulaic Structures**\n\n```\nTargets: binary contrasts, dramatic fragments, rhetorical setups, false agency\n\nTypical AI tells:\n  ✗ \"It's not about X, it's about Y\" (binary contrast)\n  ✗ \"But here's the thing.\" (dramatic standalone sentence)\n  ✗ \"So how do we...?\" (rhetorical question setup)\n  ✗ \"This opens new possibilities\" (false agency)\n\nFix: Just say Y. Cut the setup. Name who is doing what specifically.\n```\n\n**Rule 3: Active Voice**\n\n```\nEvery sentence needs a human subject performing an action.\n\n  ✗ \"This feature was designed to...\"\n  ✗ \"Research suggests...\" (research with no author)\n  ✓ \"The team designed this feature to...\"\n  ✓ \"MIT researchers found...\"\n```\n\n**Rule 4: Be Specific**\n\n```\n  ✗ \"This is a watershed moment\"\n  ✗ \"Every single person...\" / \"Always\" / \"Never\" (lazy absolutes)\n  ✓ \"[Specific event] marked [specific change]\"\n```\n\n**Rule 5: Put the Reader in the Room**\n\n```\n  ✗ \"People often find themselves...\" (narrator-from-a-distance)\n  ✓ \"You'll find...\" (direct address)\n\n  \"You\" beats \"people.\" Specific scene beats abstract description.\n```\n\n**Rule 6: Vary Rhythm**\n\n```\n  ✗ Three consecutive sentences of the same length\n  ✗ Em dashes — used like this — for emphasis\n  ✗ Things come in threes: first..., second..., third... (always three)\n  ✓ Mix long and short sentences. Two items are stronger than three.\n```\n\n**Rule 7: Trust Readers**\n\n```\nState facts directly. Skip softening and hand-holding.\n\n  ✗ \"You might be wondering...\"\n  ✗ \"Before we dive in, let me explain...\"\n  ✓ Give the information. Readers will judge it themselves.\n```\n\n**Rule 8: Cut Quotables**\n\n```\n\"If it sounds like it was written to be screenshot-shared, rewrite it.\"\n\n  ✗ \"Technology is the tool. People are the core.\"\n  ✗ \"True innovation begins with redefining the problem.\"\n\nThese sentences usually say nothing — they sound like they mean something\nbut don't actually contain information.\n```\n\nSKILL.md includes a 12-item checklist — the most practical part for daily use:\n\n| Check | Action |\n|---|---|\n| Any adverbs? | Remove them all |\n| Passive voice? | Find the actor, make them the subject |\n| Inanimate subject doing a human action? | Name the person |\n| Sentence starts with a Wh- word? | Restructure |\n| \"Here's what/this/that\" opener? | Cut to the point |\n| \"Not X, it's Y\" contrast? | Just say Y |\n| Three consecutive same-length sentences? | Break one |\n| Punchy one-liner at paragraph end? | Vary it |\n| Em-dash present? | Delete |\n| Vague declarative? | Name the specific thing |\n| Narrator-from-a-distance voice? | Place the reader in the scene |\n| Meta-joiners like \"the rest of this essay\"? | Delete; let the writing move |\n\nEach dimension scores 1–10, maximum 50. **Below 35/50: revise.**\n\n| Dimension | Core Question |\n|---|---|\nDirectness |\nMaking statements, or making announcements? |\nRhythm |\nVaried sentence lengths, or metronomic regularity? |\nTrust |\nRespects reader intelligence, or hand-holding? |\nAuthenticity |\nSounds human, or sounds AI-generated? |\nDensity |\nEvery word earning its place, or cuttable content present? |\n\nThe rubric's value: it turns \"sounds like AI\" from a vague intuition into **five specific, improvable dimensions**. When you're revising, you know what you're fixing. You also know when you can stop.\n\nThis is stop-slop's most important design insight, and what separates it from most \"AI content detection\" tools:\n\n```\nThe mainstream approach (detection):\n  AI-generated content → detector → flagged as \"AI-written\" → ...now what?\n\nstop-slop's approach (prevention):\n  Before/during writing → Skill activates → AI avoids patterns while generating\n  → Clean output, directly\n```\n\nThe difference is fundamental. Detection tools tell you \"this was written by AI\" but don't help you fix it. stop-slop is real-time interception during the writing process, not post-hoc forensics.\n\nBehind the 5.8k Stars, the more interesting number is 435 Forks — and the derivative projects they spawned:\n\nThese derivatives tell you something: **the need to identify and eliminate AI writing tells is real, persistent, and domain-specific** — every field has its own particular flavor of slop.\n\n`/plugin marketplace add hardikpandya/stop-slop`\n\nstop-slop did something that looks simple but isn't: it decomposed the vague intuition of \"AI-flavored writing\" into 8 actionable, specific rules — and with concrete rules, AI can self-correct instead of hoping for the best.\n\n*Find more useful knowledge and interesting products on my Homepage*", "url": "https://wpnews.pro/news/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to", "canonical_source": "https://dev.to/wonderlab/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to-eliminate-its-2nci", "published_at": "2026-05-28 05:40:12+00:00", "updated_at": "2026-05-28 05:53:14.727958+00:00", "lang": "en", "topics": ["generative-ai", "large-language-models", "ai-tools", "natural-language-processing", "ai-ethics"], "entities": ["stop-slop", "Hardik Pandya", "Claude Code Skill"], "alternates": {"html": "https://wpnews.pro/news/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to", "markdown": "https://wpnews.pro/news/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to.md", "text": "https://wpnews.pro/news/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to.txt", "jsonld": "https://wpnews.pro/news/one-open-source-project-a-day-no-78-stop-slop-a-skill-file-that-teaches-ai-to.jsonld"}}