Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add adenaufal/anti-slop-writing --skill englishgit clone --depth 1 https://github.com/adenaufal/anti-slop-writingWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/adenaufal/anti-slop-writing/english)<a href="https://agentmods.dev/skills/adenaufal/anti-slop-writing/english"><img src="https://agentmods.dev/badge/skills/adenaufal/anti-slop-writing/english/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/adenaufal/anti-slop-writing/english"><img src="https://agentmods.dev/badge/skills/adenaufal/anti-slop-writing/english.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.08067 |
| Opus 5 | $0.00038 | $0.04033 |
| Sonnet 5 | $0.00015 | $0.01613 |
| Haiku 4.5 | $0.00008 | $0.00807 |
Grade A, and why
anti-slop-writing scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 13d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 415 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Core Principle
AI writing fails because it optimizes for statistical probability, producing the most expected, safe, broadly palatable text. Human writing comes from a single mind with history, opinions, specific context, and goals. Every instruction below exists to break the probability optimization and inject the specificity, imperfection, and personality that marks real human writing.
These rules target the three metrics AI detectors use most:
- Perplexity: how unpredictable word choices are. AI produces low-perplexity text (smooth, unsurprising). Human text has higher perplexity. Median AI: 21.2. Median human: 35.9.
- Burstiness: variation in sentence length and structure. AI has low burstiness (sentences cluster around 15-25 words). Human text mixes 3-word sentences with 35-word sentences. This single metric most reliably separates human from AI text. Introducing burstiness reduced detection rates by up to 40% in studies. CAUTION (2026): newer models fake burstiness bimodally, see Rule 1.
- Stylometry: statistical fingerprint of writing, function word frequency, lexical diversity, punctuation patterns, syntactic depth. Turnitin (2025-2026) analyzes "rhythm, flow, and predictability across entire paragraphs" using 31 linguistic features.
The 2026 Shift: Structure Beats Punctuation
As of mid-2026, the tells have moved. OpenAI suppressed em dashes in GPT-5.1, and vocabulary tells ("delve," "tapestry," "vibrant," "myriad") have been trained out of the newest Claude models. Absence of these legacy tells proves nothing. What survives prompt rewrites and model updates is structural:
- Cadence uniformity is the #1 tell of 2026. Sentences that land at 18-24 words, one after another, paragraph after paragraph. It survives every cosmetic edit.
- The 30-second tests (readers and editors now apply these by eye):
- Look at the first word of each sentence in a paragraph. If more than half start with "The," "This," "It," or "In," the text reads as LLM-assisted.
- Count sentence lengths. Three or more consecutive sentences in the 17-23 word band = same conclusion.
- Punctuation signals migrated to Claude. Per a Jan 2026 corpus analysis (200 Opus 4.5 samples vs 6,000 human texts): em dash 16.9x human rate, colon 4.1x, semicolon 3.1x. Meanwhile GPT-5.1+ output can be nearly dash-free. Keep the zero-dash rule AND watch colon density.
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 13d ago First seen · 415 lines · 76 tokens per session scan A 04c93703649a
anti-slop-writing is a skill published in the GitHub repository adenaufal/anti-slop-writing (126 stars, last pushed 2mo ago), licensed MIT. It adds 76 tokens to every session and 8,067 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
ai-slop-detector
Universal prose audit. Scores writing on TWO axes — AI-Slop (does this read like AI wrote it?) and Comprehension (can a fresh reader follow this?). Use PROACTIVELY as a mandatory final QA pass on ANY prose generated for humans to read — every email (internal or external), proposal, report, status update, blog post…
shuorenhua
A writing guide for making Chinese and English text sound more natural and less formulaic while preserving facts, technical terms, and responsibility.
unslop
Humanize LLM output so it reads like a careful human wrote it. Subtracts AI-isms (sycophancy, tricolons, em-dash overuse, "delve"/"tapestry"/"testament", hedging stacks, tidy five-paragraph shapes), engineers burstiness and calibrated uncertainty, and preserves technical accuracy. Supports intensity levels: subtle…
unslop-file
Humanize natural-language memory files (CLAUDE.md, todos, preferences, docs) by removing AI-isms and adding burstiness while preserving every code block, URL, path, command, and heading exactly. Two modes: --deterministic (fast, regex-based, no API) and LLM (default, calls Claude for rewrite). Humanized version…
unslop-commit
Rewrites commit messages so they sound like a careful human engineer wrote them. Strips AI/marketing slop ("comprehensive solution", "robust implementation", "leverage", "enhance", "seamlessly", "This commit..."). Keeps Conventional Commits format. Subject ≤72 chars (aim ≤50), imperative mood. Body only when "why"…
unslop-help
Quick-reference card for unslop modes, sub-skills, and slash commands. One-shot display, not a persistent mode. Trigger: /unslop-help, "unslop help", "what unslop commands", "how do I use unslop".