Stats
Actions
Tags
From nullish-app
Detects mechanical "AI slop" patterns in generated text and rewrites it to sound human. Use as the final post-processing step for ANY text deliverable (blog posts, newsletters, marketing copy, business plans, proposals, reports, emails, landing-page copy). Use this skill whenever the request looks like: - "make this sound less like AI", "humanize this", "remove the AI slop" - "this reads too robotic", "make it more natural", "fix the stiff phrasing" - "this looks AI-generated", "tighten this up", "rewrite so a human wrote it" - post-processing text produced by Cowork, Claude, ChatGPT, or other AI tools Do NOT apply to: code, JSON/CSV data, charts/tables, pure numeric reports, or simple lookup answers.
How this skill is triggered — by the user, by Claude, or both
Slash command
/nullish-app:ai-slop-reviewerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> nullish-app | Text Quality
nullish-app | Text Quality
AI-generated text is usually accurate but full of predictable, mechanical patterns. This skill systematically finds those patterns and rewrites the text so the reader hears a real human voice. It runs as the LAST step of any text-producing workflow — after a draft exists, before it ships.
ai slop humanize de-ai rewrite natural tone robotic stiff post-processing polish copy editing
Forbidden words / phrases — remove or replace on sight:
Structural AI patterns:
| Pattern | Fix |
|---|---|
| Uniform sentence length (everything 15-22 words) | Mix short (3-7 words) and long (25+) |
| Every paragraph = topic -> explanation -> example -> mini-conclusion | Vary paragraph shape and length |
| AI-style intro ("In today's fast-paced world...") | Cut it; open with a concrete fact or claim |
| AI-style ending ("As we have seen...") + restating | End with a next action or a direct line |
| Over-listing (bullets for prose-worthy content) | Convert short lists to sentences |
| Passive-voice overuse | Name the actor, go active |
Return the result in this shape:
### Diagnosis
Patterns found: (e.g., AI intro, passive voice, 6 forbidden words, uniform length)
Severity: low / medium / high
### Rewritten text
(the full rewritten text)
### Key changes
(3-5 bullets describing the important edits)
For long inputs (500+ words), show the Diagnosis and Key changes first, then the rewritten text.
Example 1: business email
"Humanize this: In today's rapidly evolving landscape, it is crucial to leverage cutting-edge tools to streamline your workflow. Furthermore, it is important to note that comprehensive integration significantly enhances productivity."
Example 2: report section
"This report reads too robotic — make it sound like a person wrote it, but keep all the numbers exact."
| Output | Format | Description |
|---|---|---|
| Diagnosis | markdown | patterns found + severity (low/medium/high) |
| Rewritten text | markdown / plain | meaning preserved, human voice |
| Key changes | list | 3-5 main edits |
| Skill | Relation | Description |
|---|---|---|
| blog | before | generates blog posts that this skill polishes |
| copywriting | before | marketing copy fed into this skill |
| (any text generator) | before | this skill is the final humanization pass of the chain |
Source: ported from cowork-plugins ai-slop-reviewer (MIT), adapted to English-first detection.
npx claudepluginhub hotei09/nullish-app --plugin nullish-appGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.