Humanizer - Setup Guide¶
Source: nousresearch/hermes-agent (87 peak installs) Category: Creative / Writing License: MIT · Author: Siqi Chen (@blader), ported by Hermes Agent Version: 2.5.1 · Platforms: Linux, macOS, Windows
Identify and remove signs of AI-generated text to make writing sound natural and human. Based on Wikipedia's "Signs of AI writing" guide maintained by WikiProject AI Cleanup, derived from observations of thousands of AI-generated text instances. LLMs use statistical algorithms that tend toward the most likely completion - this skill spots and strips those telltale patterns.
What It Detects & Fixes¶
| Pattern Category | What It Catches | Fix Applied |
|---|---|---|
| Overused transitions | "Furthermore," "Moreover," "In conclusion," "Additionally" | Replaced with natural connectors or removed |
| Hedging language | "It is worth noting that," "It is important to consider" | Stripped - state the point directly |
| Forced structure | "First, ... Second, ... Finally, ..." | Converted to organic paragraph flow |
| AI vocabulary tells | "delve," "tapestry," "landscape," "realm," "crucial," "paramount" | Replaced with plain-language alternatives |
| Overly formal tone | Academic register when casual is appropriate | Adjusted to match target voice |
| Repetitive sentence rhythm | Uniform sentence length and structure | Varied for natural cadence |
| Fake empathy | "I understand your frustration," "That's a great question" | Removed unless genuinely warranted |
Installation¶
Via skills.sh (Recommended)¶
npx skills add nousresearch/hermes-agent --skill humanizer
Direct from Hermes Agent Repo¶
# Clone and copy the skill
git clone --depth 1 https://github.com/nousresearch/hermes-agent.git /tmp/hermes-agent
cp -r /tmp/hermes-agent/skills/creative/humanizer ~/.hermes/skills/
Manual (Single Install)¶
# Create the skill directory
mkdir -p ~/.hermes/skills/humanizer
# Download SKILL.md
curl -o ~/.hermes/skills/humanizer/SKILL.md \
https://raw.githubusercontent.com/nousresearch/hermes-agent/main/skills/creative/humanizer/SKILL.md
Usage with Hermes Agent¶
Trigger Phrases¶
The skill auto-loads when you ask Hermes to: - "humanize this text" - "de-AI this paragraph" - "make this sound less like ChatGPT" - "strip the AI-slop from this" - "rewrite this in my voice"
Input Methods¶
| Method | How | Best For |
|---|---|---|
| Inline | Paste text directly in chat | Quick edits, short passages |
| File | Point at a file path | Documents, blog posts, PR descriptions |
| Voice calibration | Provide a writing sample | Matching your specific voice |
Voice Calibration¶
For best results matching your personal voice, provide a sample of your own writing:
"Humanize this blog post, matching the voice in ~/writing-samples/my-style.md"
The skill reads your sample first, extracts voice characteristics, then rewrites accordingly.
Quick Example¶
Input (AI-written):
"Furthermore, it is worth noting that the implementation of this innovative solution will undoubtedly revolutionize the landscape of modern computing paradigms."
Output (humanized):
"This solution changes how we think about computing."
Verification¶
After installation, verify the skill is loaded:
hermes skills list | grep humanizer
Test in a Hermes session:
You: Humanize this: "It is crucial to understand that leveraging AI capabilities can significantly enhance organizational productivity metrics."
Hermes: [Returns de-slopped version]
Related Skills¶
- songwriting-and-ai-music - AI music generation with similar creative voice control
- ultimate-humanizer - Community alternative with 50 patterns and 5D scoring
- research-paper-writing - Academic writing with the same anti-AI-slop pass
Pro Tips¶
- Run a pre-publish pass - Load humanizer before sending any user-facing text (release notes, docs, PR descriptions).
- Combine with voice calibration - The 5-minute investment of providing a writing sample pays off in every subsequent humanization.
- Not just for English - The patterns are language-agnostic; the skill works across any language the underlying model supports.
- Iterative refinement - Run the humanizer, read the result, then ask for specific adjustments ("make it more casual," "shorter sentences," "add more personality").
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.