The anti-slop editor for AI writing.
Less slop, more pop. No rewrites.
An open-source skill with an agentic pipeline and a memory. It finds and fixes slop in anything you write with AI: LinkedIn and social posts, emails, blogs, documents, slides, research papers. Say no to AI slop.
Free · open source · MIT
npx skills add manavmishra/ZeroSlop --globalor paste this prompt into your favorite agent Claude Code · Codex · Claude Cowork · any harness
Install the Zero Slop skill by running npx skills add manavmishra/ZeroSlop --global, then use it to edit my drafts.Every number comes from the public repository: a blind two-rater quality study, an 18-draft replay across five workflows, and pinned data files. Source snapshot 43299df
From LinkedIn posts to research papers: tasteful rewrites with zero slop.
We’re thrilled to announce that our team has leveraged cutting-edge AI to deliver a seamless onboarding experience, cutting setup time by 40%. This milestone underscores our unwavering commitment to innovation and operational excellence.
We cut onboarding setup time by 40% using AI.
In today’s rapidly evolving landscape, AI is enabling content teams to accelerate the drafting process while maintaining authenticity. In this article, we’ll explore how thoughtful review workflows can bridge the gap and empower writers to create meaningful content.
AI helps content teams draft faster without losing their voice. This article looks at review workflows that protect authenticity and help writers produce work that matters.
To unlock scalable value, the organization should leverage a comprehensive, cross-functional transformation framework that aligns AI-enabled operating models with strategic priorities.
This integrated approach will empower teams to optimize decision-making and drive sustainable impact across the enterprise.
The organization should use one cross-functional plan to align its AI operating models with its priorities.
That plan should help teams make better decisions and create lasting value.
AI is transforming how we write, but here’s what nobody tells you: the future isn’t about replacing writers.
It’s about leveraging AI to accelerate the drafting process while unlocking the unique human creativity that makes great writing possible. A thread 🧵
AI can speed up drafting without replacing writers.
Human creativity still makes great writing possible.
The workflow scores the sloppiness of your draft, uses your AI model to edit it in context, then runs the edited text through a verifier loop.
Separate diagnostics: a host-model predictability probe, and a portfolio probe for repetition across 3+ related drafts. Neither changes the writing score.
A cut span, a kept flag, a preferred fix. What you keep becomes evidence.
Recurrence, novelty, and known-human safety. One edit cannot activate a rule.
Recurring fixes guide later edits. Stale guidance decays.
This does not retrain the host model or perform reinforcement learning or RLHF. It updates an inspectable private detector and rewrite memory.
The score is built from four inputs, each tracing back to a phrase or a statistic. A high number indicates AI slop.
Pattern meter
267 weighted patterns and a 96-word watchlist, plus 25 context-gated terms: the same word can be ordinary vocabulary in one sentence and a tell in the next.
Rhythm
Followability
Comma pileups, long-word clusters, and sentences past 38 words all get flagged.
Formatting and register
Emoji strings, hashtag clusters, heavy bolding, repeated dashes, machine-formal language. None of them is damning alone, but they turn suspect when they cluster.
Your edits teach Zero Slop.
Every edit you keep or reject is a hint: a missed tell, or a false alarm.
A phrase or a single word must show up in multiple separate edits before it signals slop.
Learning stays private to you, and it sharpens both detection and fixes.
72 method-hidden passages, two blind LLM raters, one rubric. Then an 18-draft replay across five editing workflows. The repository includes the data, the labels, and the limits.
Read the benchmark notesThe public bench.
30,000We scored 18/18 on our own sample corpus. Then we ran it against almost 30,000 pieces of writing, other people’s benchmarks included, and you can look at all results.
Other editors are a prompt. Zero Slop delivers an anti-slop learning agentic workflow. We audited Zero Slop with two top skills used today: no-ai-slop and humanizer. Supported means real code ships the check. Partially supported means a prompt asks the model and trusts it.
| Capability | Zero Slop | blader/humanizer | petergyang/no-ai-slop |
|---|---|---|---|
| Score a draft without rewriting it | Supported | Not supported | Partially supported |
| A slop score you can trace to each phrase | Supported | Not supported | Not supported |
| Rhythm and formatting checks | Supported | Partially supported | Partially supported |
| A script that checks every fact | Supported | Partially supported | Partially supported |
| Writes several rewrites, keeps the best | Supported | Not supported | Not supported |
| Separate grammar and read-aloud passes | Supported | Partially supported | Partially supported |
| Catches repetition across your drafts | Supported | Not supported | Not supported |
| Learns from your edits, privately | Supported | Not supported | Not supported |
| Remembers the fixes you keep | Supported | Not supported | Not supported |
| Re-checks what it learned; stale rules expire | Supported | Not supported | Not supported |
| Learns your voice from a sample | Supported | Partially supported | Not supported |
| A public test suite for its results | Supported | Not supported | Not supported |
We audited each complete repository at a pinned commit. Supported means real code ships the check: a dedicated script, stored artifact, or named workflow gate. Partially supported means a prompt asks the model, with nothing enforcing it. Not supported means the capability is absent from that repository at the audited commit.
A documented capability is not proof of better writing.
Open the audit dataA plausible-looking draft saves the sender a few minutes and hands the reader a harder job. It also reduces trust and credibility besides burning productivity: in the September 2025 BetterUp and Stanford survey, 42% of recipients saw the sender of workslop as less trustworthy, about half saw them as less capable or reliable, and nearly one in three would avoid working with them again.
What workers report
Workers who got sloppy AI work said it took nearly two hours to fix each piece. Four in ten had received some in the past month. BetterUp and Stanford study
From a controlled study
In METR's 2025 experiment, experienced developers expected AI to speed them up. It slowed them down by 19%. One narrow setting, not a verdict on every developer or newer tool. METR study
From workplace data
Gallup's May 2026 data shows the gap: companies add AI tools faster than employees pick them up, and a clear plan is rarer still. Gallup indicator
Among the 31,476 developers who answered that survey question, 45% also said debugging AI-generated code takes more time. The setting is software, but the failure mode is familiar: fluent output still needs expert verification. Stack Overflow survey
In the BetterUp and Stanford survey, the cleanup shifted effort across the team and changed how recipients viewed the colleague who sent it. Survey source
RAND cites external estimates of more than 80%. Its own interviews traced failure to problem definition, workflow fit, data, infrastructure, and technology-first planning. Sloppy output can add review debt, but it is not the whole explanation. RAND report
Zero Slop judges clusters of evidence, never a single signal. One em dash does not make a draft machine-written, and formal does not mean fake.
Language changes; claims do not. No faked warmth, no invented specifics.
Figures, names, quotes, and links all verified before a rewrite passes.
A sample of your writing teaches the scorer what is simply you.
Zero Slop plugs into any agent that reads SKILL.md files: Codex, Claude Code, Cursor, Gemini CLI, OpenCode, Warp, and Zed. Open-source favorites like VS Code and Goose support the same standard.
Ask your agent to edit your writing with Zero Slop.
Free · open source · MIT
npx skills add manavmishra/ZeroSlop --globalor paste this prompt into your favorite agent Claude Code · Codex · Claude Cowork · any harness
Install the Zero Slop skill by running npx skills add manavmishra/ZeroSlop --global, then use it to edit my drafts.Zero Slop's scoring, rewrite, and copy-desk instructions
Compatible agents
What the tool does, where it runs, and what it costs.
AI slop is writing that falls into familiar model habits: safe phrasing, even rhythm, template structure, stock transitions, and polished sentences that say very little. Zero Slop measures those surface patterns and shows you exactly where they occur.
No. A detector guesses whether a machine wrote something. Zero Slop scores the writing itself, lists every flagged phrase, edits the draft, and checks figures, names, quotes, and links against the original. The score does not identify the author.
No. Zero Slop is an anti-slop learning agentic workflow. Seven specialists, scripts and AI, pass your draft along one pipeline; a verifier rescores the result; a private loop learns from the edits you keep. A prompt instructs once. This workflow measures, edits, verifies, and learns.
Zero Slop is designed to prevent that. It inventories figures, names, quotes, and links, then rejects a rewrite that drops one or adds a new one. You still make the final editorial call.
No. The scorer is a local Python script that uses only the standard library. It works offline and needs no account, server, or network connection. The skill itself runs inside the coding or writing agent you already use.
Zero Slop works with agents that support SKILL.md files, including Codex, Claude Code, Cursor, Gemini CLI, OpenCode, Warp, and Zed. The repository also documents setup for ChatGPT and claude.ai.
Yes. Zero Slop is open source under the MIT license. The scorer uses only Python's standard library; it needs no third-party packages.
Zero Slop can compare the draft it returned with the version you publish. One edit cannot change the meter. A phrase must recur across 3 separate edits; a single word across 5. The evidence must also be new, and it must never flag the known-human reference set. What passes stays in a private local overlay and shapes the next run.
The website records aggregate page traffic, performance, chapter views, scroll depth, example choices, and clicks on install, GitHub, feedback, and sharing controls. No cookies, persistent identifiers, fingerprinting, or draft text are collected. Global Privacy Control and Do Not Track disable the custom interaction events.