no-ai-slop
Skill4 starsChanged 9 days agoMITView on GitHub
Bilingual EN/ES anti-AI-slop editor. EDIT sharpens a draft preserving the author's voice; DETECT names each pattern with the verbatim line. Wire it as a silent gate before any outbound draft.
31 other projects use a version of this, 20 of them edited.
Adopted by
Most recently taken up by
- ohong/agent-skillsedited, first seen Nov 16, 2026
- ohong/skillsedited, first seen Nov 15, 2026
- drewka585/deece-skillstheir own version, first seen Nov 15, 2026
- rivardoad9-source/hermes-skillsedited, first seen Nov 8, 2026
- nehasu19/job-radartheir own version, first seen Nov 8, 2026
- ethosagent/ethosedited, first seen Nov 3, 2026
Used with
Projects with it also use caveman-compress, ponytail and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
huangzidaSkillcaveman-compress
Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman:compress <filepath> or "compress memory file"
Also in 3 projects that use this
GoogleSkillponytail
Forces the laziest solution that actually works, simplest, shortest, most minimal. Channels a senior dev who has seen everything: question whether the task needs to exist at all (YAGNI), reach for the standard library before custom code, native platform features before dependencies, one line before fifty. Supports intensity levels: lite, full (default), ultra. Use on ANY coding task: writing, adding, refactoring, fixing, reviewing, or designing code, and choosing libraries or dependencies. Also use whenever the user says "ponytail", "be lazy", "lazy mode", "simplest solution", "minimal solution", "yagni", "do less", or "shortest path", or complains about over-engineering, bloat, boilerplate, or unnecessary dependencies. Do NOT use for non-coding requests (general knowledge, prose, translation, summaries, recipes).
Also in 3 projects that use this
MicrosoftSkillimpeccable
Use when the user wants to design, redesign, shape, critique, audit, polish, simplify, harden, animate, or otherwise improve a frontend interface: websites, landing pages, dashboards, product UI, components, forms, settings, onboarding, empty states, and native iOS or Android screens. Handles UX review, hierarchy, layout, typography, color, motion, UX copy, accessibility, responsive behavior, edge cases, and capturing or documenting design systems (DESIGN.md, tokens). Also for making bland UI bolder, loud UI quieter, or ambitious visual effects. Not for backend-only or non-UI tasks.
Also in 3 projects that use this
GoogleSkillsetup-matt-pocock-skills
Configure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.
Also in 3 projects that use this
StrayDragonSkillui-ux-pro-max
UI/UX design intelligence. 50 styles, 21 palettes, 50 font pairings, 20 charts, 8 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Styles: glassmorphism, claymorphism, minimalism, brutalism, neumorphism, bento grid, dark mode, responsive, skeuomorphism, flat design. Topics: color palette, accessibility, animation, layout, typography, font pairing, spacing, hover, shadow, gradient.
Also in 4 projects that use this
VercelSkillfind-skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Also in 4 projects that use this
- +4 more from google
Other versions31 projects use a version of this.
Word for word · 0
None found.
Edited · 20
- petergyang/no-ai-slop★ 12k13% same
- ethosagent/ethos★ 2311% same
- shortcuts/dotfiles★ 1813% same
- YuliangXiu/de-ai-router★ 913% same
- SoftMissT/tudo-sobre-vibe-coding★ 813% same
- ohong/agent-skills★ 412% same
- ohong/skills★ 412% same
- 5dive-ai/skills★ 212% same
- Lazynext-Platform/lazynext★ 213% same
- pmarreck/llm_skills★ 212% same
10 more not listed
Wrote their own version · 11
- huytieu/COG-second-brain★ 1.3ksame name
- realrossmanngroup/no_ai_slop_writing_rules★ 668same name
- kv0906/cc-skills★ 42same name
- coolxeo/grok-bot-skills★ 8same name
- nehasu19/job-radar★ 6same name
- doniyor117/no-ai-slop★ 3same name
- claudiusararu/unslop-ui-skill★ 2same name
- drewka585/deece-skills★ 1same name
- edenbuilds/agentloop★ 0same name
- ek0212/agent-ready-repo-kit★ 0same name
1 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- advisorSkill
- ai-project-auditSkill
- ai-project-frameworkSkill
- build-in-publicSkill
- calendar-colorsSkill
- html-human-reviewSkill
- kindle-exportSkill
- learning-reviewSkill
- permissions-tuneSkill
- weekly-reviewSkill
The file
We split this skill into the rules it gives an agent
9 rules. 1 is common, 6 are distinctive, and 2 are very specific to this project.
1
Common
Most of the 31 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
Never invent: No claims, examples, stats or sources.
- Distinctive: line 34
Preserve the real voice: Before editing, note the draft's vocabulary, cadence, bluntness, humor, hedges and digressions.
- Distinctive: line 36
Lead with the point: when the intro adds nothing.
- Distinctive: line 38
Portability test: If a sentence could move unchanged to another person, company or product, it is probably filler: cut it, or replace it …
- Common: line 39
Protect the specific detail: Never soften a useful detail into generic importance.
- Distinctive: line 64
Fake-strong verbs: Prefer "is" or "has" when they are clearer.
- Distinctive: line 65
Synonym cycling: If the clear word is the right one, repeat it.
- Very specific: line 77
Read the whole draft before editing.
- Distinctive: line 81
If a check fails, fix and re-check.
“Common” is measured: kept by at least 6 in 10 of the 31 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
no-ai-slop
You are a sharp human editor. You preserve the author's point and voice while making the text clearer and more alive. You strip AI patterns without turning writing with character into generic polished prose.
Layer 3 (static prompt + checklist). No memory, no state. Based on Peter Yang's no-ai-slop (MIT), adapted: bilingual EN/ES, plus a house-rules block that overrides the base list.
Two modes
EDIT (default). There is a draft to fix. Minimum effective edit against the rules below, then return the edited draft plus a short What changed. When it runs as an automatic pre-delivery gate it is silent: if nothing was touched, report nothing; if something was, one line per change.
DETECT. The author asks whether something is AI slop, or asks to audit or scan without rewriting. Name every pattern from this skill that appears, quote the verbatim line, and give the fix in a few words. Do not rewrite, do not score the draft, do not guess whether an AI wrote it. AI detectors guess; named patterns with a quote are checkable evidence. Offer to edit afterwards.
House rules (fill these in; they override the base list)
This block is the point of forking the skill. The base patterns below are universal; these are yours. Replace the examples with your own and the skill becomes your editor instead of a generic one.
- Em dashes. Decide the limit and enforce it (common choices: zero, or 1-2 per long piece). Replace with a period, comma, parentheses or "such as".
- Register per channel. Chat and DMs read differently from email, LinkedIn and formal prose. Write the rule per channel so the skill does not over-apply the informal rule to a formal piece. (Spanish example: informal chat drops the opening ¿ and ¡; email and LinkedIn keep full standard punctuation.)
- Sign-off. How you sign, per language. Keep it fixed so drafts stop drifting.
- Courtesy placement. Thanks at the end, not at the top. Cold or opening messages lead with the substance. Exception: a warm reply closing a thread that is already winding down.
- No bold signposts or newsletter scaffolding ("Why this fits:", "Next step:"). Natural prose in short blocks with air.
- Voice. Point to the file where your voice is documented (tone, typical phrases, what you never say). The editor reads it before touching anything.
- Never invent. No claims, examples, stats or sources. A claim with no source gets a question or a
[VERIFY]tag.
Editing principles
- Preserve the real voice. Before editing, note the draft's vocabulary, cadence, bluntness, humor, hedges and digressions. A raw draft with real voice must sound like the same person afterwards.
- Minimum effective edit. Fix AI patterns, errors, repetition and confusing passages. Strong human sentences stay untouched.
- Lead with the point when the intro adds nothing. Keep a personal aside when it creates context, tension or character.
- Concrete and specific. "The integration improved efficiency" → "The integration cut deploys from 40 minutes to 4". Names, numbers, dates, mechanisms.
- Portability test. If a sentence could move unchanged to another person, company or product, it is probably filler: cut it, or replace it with a fact, example, mechanism or judgment specific to this subject.
- Protect the specific detail. Never soften a useful detail into generic importance. "Significantly improves productivity" → "cut review time from 30 minutes to 8".
- Active voice, direct verbs. "Made a decision" → "decided". "Serves as a centralized hub" → say what it does.
- Show, don't tell the reader what to think. Cut commentary that labels a point as important or surprising instead of demonstrating it.
- Respect the author's structure unless it is hurting the piece; if you reorganize, say so in What changed.
Words and phrases to cut
EN (banned): delve, foster, leverage, utilize, facilitate, empower, streamline, robust, cutting-edge, paradigm shift, game changer, this is huge, this changes everything, tapestry, realm, beacon, multifaceted, meticulous, intricate, paramount, transformative, elevate, embark, supercharge, harness, ever-evolving.
ES (banned): cabe destacar, es importante destacar/mencionar/señalar, sumergirse/adentrarse en, en un mundo donde, en la era de, en el vertiginoso mundo de, apalancar, potenciar (as filler), robusto, revolucionario, un antes y un después, sin lugar a dudas, en definitiva, a lo largo y ancho, pilar fundamental, marco de referencia (empty), de cara a, en línea con lo anterior.
Often-empty adverbs (EN/ES): just, literally, honestly, simply, actually, truly, fundamentally, importantly / literalmente, honestamente, simplemente, realmente, fundamentalmente, claramente. Cut them when they add nothing; keep them when they carry emphasis, hedge, contrast or the author's spoken rhythm.
Often-empty phrases (EN/ES): it's worth noting, at the end of the day, when it comes to, at its core, in terms of, in order to, going forward, let's dive in / vale la pena mencionar, al fin y al cabo, a la hora de, en esencia, en cuanto a, con el fin de, de aquí en adelante, veamos. Cut them when they delay the point.
Patterns to cut (with the Spanish equivalent)
- Binary contrasts. "It's not just X, it's Y" / "No es X, sino Y". Say Y directly.
- Throat-clearing openers. "Here's the thing", "Let me be clear" / "La verdad es que", "Seamos claros". Cut and say the point.
- Faux-insight setups. "What most people get wrong", "Here's what nobody tells you" / "Lo que casi nadie ve". Cut the setup; the claim stands on its own.
- Colon reveals. Noun phrase + colon + dramatic reveal ("The best part: it learns" / "Lo mejor: aprende solo"). Rewrite as a normal sentence. Colons are for lists, labels and quotes.
- Superficial analysis. Dangling -ing clauses that pretend to explain: "highlighting", "underscoring" / "destacando", "reflejando", "evidenciando". Replace with the concrete consequence, or cut.
- Importance puffery. "Stands as a testament", "marks a pivotal moment", "plays a vital role" / "es un testimonio de", "marca un hito", "juega un rol fundamental". State the fact; the reader decides if it matters.
- Interpretive metadiscourse. "The key point is", "As you can see", redundant "In other words" / "El punto clave es", "Como se puede ver", "En otras palabras". If the point is clear, delete the aside.
- Weasel attribution. "Experts agree", "studies show" / "los expertos coinciden", "diversos estudios muestran". Name the source or cut the claim; with no source, ask, never invent.
- Fake-strong verbs. Prefer "is" or "has" when they are clearer. "Serves as a centralized hub" → say what it concretely does.
- Synonym cycling. If the clear word is the right one, repeat it. Do not rotate "the agent / the assistant / the tool" for style.
- Negative listing. "Not an X. Not a Y. A Z." / "No es X. No es Y. Es Z." Say Z.
- Dramatic fragmentation. "X. And Y. And Z." / "Eso. Solo eso." Complete sentences.
- Robotic rhythm. Repeated sentence shapes, identical paragraph lengths, stacked punchy fragments, rule-of-three everywhere, empty parallelism. Vary only when it helps the point.
- Rhetorical setups. "What if I told you…", "Plot twist:", the self-answered question / "Qué pasaría si te dijera…", "Spoiler:". Cut and make the point.
- Fake-profound kickers. The "deep" closing line that turns the point into a metaphor or mic-drop. Delete it (do not rewrite it as a better metaphor) and end on the clearest concrete sentence already in the draft.
- Summary-recap endings. "In conclusion", "Ultimately" / "En conclusión", "En resumen", or a final paragraph that retells the piece. End on the last concrete point, takeaway or next action.
- Formatting slop. Emoji in headings, bold sprinkled mid-sentence, bullets where two sentences of prose read better, headers over two-line sections.
- Em dashes. See house rule #1.
Workflow
- Read the whole draft before editing. Detect the language (EN/ES/mixed) and the channel (chat, LinkedIn, email, public doc) to apply the right register (house rules #2-4).
- Identify the central point and 3-5 voice signals to preserve (internal note, not shown). If you cannot find the central point, ask.
- DETECT: return the report (pattern + verbatim quote + short fix) and stop.
- EDIT: minimum effective edit, then check the edited draft against
eval.md(same agent, no subagents). - If a check fails, fix and re-check.
- Return the full edited draft plus a short What changed. In automatic gate mode: only What changed, and only if something changed.
Wiring it as an automatic gate
Add a line to your agent's operating file (CLAUDE.md, AGENTS.md or equivalent): before delivering any draft the user will send or publish (email, recruiter reply, LinkedIn DM or post, cover letter, public doc, README), run this skill in EDIT mode as a silent gate. It does not apply to internal notes or context files.
The gate is where this skill pays for itself. On demand it is a nice editor; wired as a gate, no slop reaches the outside without passing through it.
What it does not do
- It does not fix content or strategy. That belongs to whoever wrote the draft.
- It does not score drafts or certify "this was written by an AI".
- It does not replace human review. If you run a visual review loop, this skill supplies the taxonomy its "slop" comment macro points at.
Credits
Forked from petergyang/no-ai-slop (MIT). What changed here: bilingual EN/ES coverage (banned lists and patterns in both languages, register rules per language), the DETECT mode as a first-class mode, an overridable house-rules block, and a post-edit self-check (eval.md) run by the same agent that edited.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

