humanize-text — Claude Code skill: score text, strip AI patterns
humanize-text grades a piece of writing on seven scales from 1 to 10 and returns a rewritten version without the usual AI tells.
MIT 214 installs ★ 35 updated 2026-10-21
npx skills add podo/design-agent-skills --skill humanize-text -a claude-codeWhat humanize-text does
Give it a block of text, a screenshot of an interface, or a Figma link. The material gets scored across seven dimensions, each from 1 to 10, and comes back rewritten.
The dimensions: worn-out vocabulary, padding that carries no information, grammar habits, how well interface copy reads, structure, punctuation, and sentences that talk about the text itself. The checklist follows the Signs of AI Writing guide on Wikipedia. Figma links are read through the Figma MCP server. Plain text and screenshots skip it, and a design link will not resolve without the server running.
It works on copy that already exists. Fresh interface strings belong to ux-writing-skill, marketing pages to copywriting-skill, and a full design pass to design-auditor or plan-design-review.
What it helps with
- strip AI tells from a draft before publishing
- score button labels and error messages
- rewrite a paragraph bloated with filler transitions
- review copy sitting on a Figma frame
- catch em-dash overuse and stiff phrasing
Example prompts for the agent
- Score
landing.mdon all seven dimensions and show me the rewrite. - Run
humanize-textoncheckout.pngand rewrite the button labels. - Check this paragraph for AI patterns and give me a cleaner version.
How to install humanize-text
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill humanize-text -a claude-code2Manually: put the files in place
~/.claude/skills/humanize-text/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/humanize-text/
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill humanize-text -a codex2Manually: put the files in place
.agents/skills/humanize-text/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill humanize-text -a cursor2Manually: put the files in place
.cursor/skills/humanize-text/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install podo/design-agent-skills skills/humanize-text2Manually: put the files in place
.github/skills/humanize-text/SKILL.md1Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill humanize-text -a gemini-cli2Platform CLI
gemini skills install https://github.com/podo/design-agent-skills --path skills/humanize-textManually: put the files in place
.devin/skills/humanize-text/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add podo/design-agent-skills --skill humanize-text -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about humanize-text
How do I install humanize-text in Claude Code?
Run npx skills add podo/design-agent-skills --skill humanize-text -a claude-code. The skill file lands at ~/.claude/skills/humanize-text/SKILL.md.
What does humanize-text do?
It rates a text on seven scales from 1 to 10 and returns a human rewrite that drops the familiar AI writing habits.
Does the skill need a Figma MCP server?
Only when you hand it a design link. Raw text and screenshots are handled without one.
Source text
SKILL.md · MIT
---
name: humanize-text
description: |
Scores and rewrites text to remove AI writing patterns. Rates content across
seven dimensions (AI vocabulary, content inflation, grammar patterns, UX copy
quality, structural tells, punctuation, and meta-content), each on a 1–10
scale. Works with raw text, Figma URLs, and UI screenshots. Based on
Wikipedia's Signs of AI Writing field guide.
triggers:
- "humanize text"
- "remove AI patterns"
- "AI slop"
- "de-slop"
- "AI writing detection"
- "UX copy quality score"
- "humanize-text"
das:
type: skill
category: visual-components
upstream: "https://github.com/gregorymm/humanize-text"
upstream_path: "skills/humanize-text/SKILL.md"
version: latest
install: true
---
# humanize-text
> Catalogue stub — full skill: [gregorymm/humanize-text](https://github.com/gregorymm/humanize-text)
<!-- das:install:start — generated from stub.yaml by `npm run generate`; edit stub.yaml, not this block -->
## Decision tree
1. **Is the full skill already installed?**
- Global: `f=~/.agents/skills/humanize-text/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"`
- Project: `f=.agents/skills/humanize-text/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"`
- `installed` → the skill is installed as `humanize-text`, invoke it and proceed
- `pointer` → go to step 2
2. **Detect scope, then install:**
```bash
[ -e ~/.agents/skills/humanize-text ] && echo "global" || echo "project"
```
**Global:**
```bash
npx skills add gregorymm/humanize-text/skills/humanize-text -g -y
```
**Project:**
```bash
npx skills add gregorymm/humanize-text/skills/humanize-text -y
```
> **Claude Code:** send a command as a chat message starting with `!` to run it without leaving the conversation.
**Claude Code plugin route:**
```
/plugin marketplace add gregorymm/humanize-text
/plugin install humanize-text@gregorymm-humanize-text
```
**Heads-up**
- **Requires:** Figma MCP server (only for the Figma-URL input; raw text and screenshots need nothing)
- **License:** MIT
<!-- das:install:end -->
## Invoke after install
- Skill name: `humanize-text`
- Trigger phrases: "humanize text", "remove AI patterns", "de-slop", "UX copy quality score"
## What it does
Analyzes any text and produces a per-dimension score (1–10) plus a human-rewritten version:
| Dimension | What it catches |
|---|---|
| **AI vocabulary** | Overused words: "delve", "leverage", "utilize", "seamlessly", "robust" |
| **Content inflation** | Unnecessary padding, redundant qualifiers, filler transitions |
| **Grammar patterns** | Passive overuse, nominalization, non-human sentence openers |
| **UX copy quality** | Button/label clarity, error message tone, empty state effectiveness |
| **Structural tells** | Excessive bullet lists, formulaic H2/H3 patterns, intro-body-conclusion rigidity |
| **Punctuation** | Em-dash overuse, ellipsis as hedge, unnatural comma placement |
| **Meta-content** | Self-referential phrases ("In this article…", "It's important to note…") |
Works across input types: paste raw text, share a UI screenshot, or provide a Figma URL (the Figma workflow reads the design through the Figma MCP server).
## When NOT to use
- Writing new UI copy from scratch — use `ux-writing-skill`
- Writing new marketing copy — use `copywriting-skill`
- Full design audit — use `design-auditor` or `plan-design-review`
© podo · MIT · podo/design-agent-skills