Humanize Ai Text

Detect and remove AI-generated writing patterns to rewrite text for a more natural, human-written style with diagnostic reports and version comparison.

Security Scan

The skill's instructions claim runnable Python scripts and local pattern files, but the package contains only SKILL.md and README — the executable components are missing, which is an incoherence that warrants caution before installing or running anything.

Purpose & Capability

The name and description (detect and rewrite AI-like text) match the claimed behavior, but the SKILL.md repeatedly references Python scripts (detect.py, transform.py, compare.py) and a patterns.json that are not present in the file manifest. The skill also declares Python3 is required in metadata even though no scripts are included—this mismatch suggests the distributed bundle is incomplete or mispackaged.

Instruction Scope

Runtime instructions tell the agent to execute local Python scripts, edit patterns.json, and process arbitrary files — actions that require code that isn't bundled. While the described actions (analyze and rewrite files) are within the stated purpose, the SKILL.md gives explicit commands that would run arbitrary Python code if those scripts existed. The instructions also assert 'No external network calls' and 'No credentials' but those are claims in prose and cannot be verified because the referenced scripts are missing.

Install Mechanism

There is no install spec and no code will be written to disk by an installer (instruction-only), which is low installation risk. However, the skill claims the tooling is 'already installed' and depends on Python scripts that are absent—this is an integrity/packaging issue rather than an install-method risk.

Credentials

The skill requests no environment variables or credentials, which is proportional to a text-processing tool. There is no evidence of requests for unrelated secrets or config paths in the provided files.

Persistence & Privilege

The skill does not request persistent/always-on privileges and uses default autonomous invocation settings. There is no indication it attempts to modify other skills or system-wide agent settings in the provided files.

What to consider before installing

Do not assume the skill is usable as-is. SKILL.md instructs running Python scripts (detect.py, transform.py, compare.py) and editing patterns.json, but the package contains only README.md and SKILL.md — the executable scripts are missing. This can mean a packaging error, an incomplete upload, or that the real code is fetched from elsewhere (not shown). Before installing or running anything: 1) Request the missing source files from the publisher or a full repository link; 2) If you obtain the scripts, review them (or run them in an isolated sandbox) to verify there are truly no network calls or secret access; 3) Prefer skills with a verifiable source (homepage or repo) and explicit install artifacts; 4) Be cautious about running arbitrary Python scripts from unknown authors even if they claim 'no network calls.' If you cannot verify the code, do not grant this skill execution privileges or include it in automated pipelines.

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written.

Based on Wikipedia's comprehensive "Signs of AI writing" guide.

Features

• AI Detection: Identifies 16 pattern categories of AI-generated text

• Pattern Categories:

  • Citation bugs
  • Knowledge-cutoff phrases
  • Chatbot artifacts
  • Markdown/code remnants
  • AI-specific vocabulary
  • Filler phrases
  • Punctuation quirks
  • Stylistic parallelisms
  • Inflated symbolism
  • Promotional language
  • Superficial -ing analyses
  • Vague attributions
  • Em dash overuse
  • Rule of three
  • Negative parallelisms
  • Excessive conjunctive phrases

• Diagnostic Reports: Generate detailed AI detection reports

• Text Transformation: Rewrite text to sound more human

• Comparison Mode: Compare original vs humanized versions

Usage

Detect AI Writing

# Basic detection
python skills/humanize-ai-text/scripts/detect.py text.txt

# JSON output
python skills/humanize-ai-text/scripts/detect.py text.txt -j

# Score only
python skills/humanize-ai-text/scripts/detect.py text.txt -s

# Pipe text directly
echo "This text might be AI-generated" | python skills/humanize-ai-text/scripts/detect.py

Transform Text

# Basic transformation
python skills/humanize-ai-text/scripts/transform.py text.txt

# Output to file
python skills/humanize-ai-text/scripts/transform.py text.txt -o clean.txt

# Aggressive mode (more changes)
python skills/humanize-ai-text/scripts/transform.py text.txt -a

# Quiet mode (no progress output)
python skills/humanize-ai-text/scripts/transform.py text.txt -q

Compare Versions

# Compare original and transformed
python skills/humanize-ai-text/scripts/compare.py text.txt

# Compare and save result
python skills/humanize-ai-text/scripts/compare.py text.txt -o document_v2.txt

Installation

The skill is already installed. To use it, ensure you have Python 3 installed.

Dependencies

No external dependencies required - uses Python standard library.

Pattern Detection

The detection script identifies these AI writing patterns:

1. Citation Bugs

• Fake or unverifiable sources
• "According to experts" without specifics
• Generic study references

2. Knowledge Cutoff

• "As of my last update"
• "I don't have real-time information"
• Date-stamped limitations

3. Chatbot Artifacts

• "I'd be happy to help"
• "Great question!"
• "Let me break this down"
• Excessive hedging

4. AI Vocabulary

• "delve into"
• "testament to"
• "in conclusion"
• "it's worth noting"
• "important to understand"

5. Filler Phrases

• Unnecessary introductions
• Redundant summaries
• Padding content

6. Stylistic Issues

• Rule of three overuse
• Negative parallelisms
• Em dash overuse
• Excessive conjunctive phrases

Best Practices

1. Detect First: Run detection before transformation
2. Review Changes: Always review transformed text
3. Use Compare: Compare versions to understand changes
4. Iterate: May need multiple passes for best results
5. Custom Patterns: Edit patterns.json to add custom detections

Customization

Add Custom Patterns

Edit skills/humanize-ai-text/scripts/patterns.json:

{
  "custom_patterns": {
    "my_pattern": {
      "regex": "pattern_here",
      "replacement": "replacement_text",
      "description": "What this pattern detects"
    }
  }
}

Batch Processing

Process multiple files:

for f in *.md; do
  python skills/humanize-ai-text/scripts/detect.py "$f" -s
  python skills/humanize-ai-text/scripts/transform.py "$f" -a -o "${f%.md}_clean.md" -q
done

Output Formats

Detection Output

AI Detection Report
===================
File: text.txt
AI Probability: 73%

Patterns Detected:
- AI Vocabulary: 12 instances
- Filler Phrases: 8 instances
- Stylistic Parallelisms: 5 instances
...

Transformation Output

Original: "I'd be happy to delve into this important topic."
Humanized: "Let's explore this topic."

Security Notes

• ✅ No external network calls
• ✅ All processing done locally
• ✅ No credential access
• ✅ Safe file operations only
• ✅ No data exfiltration

Integration

With Writing Workflow

1. Write draft (AI-assisted or not)
2. Run detection to check AI score
3. Transform if score is high
4. Compare versions
5. Manual review and final edits

With Content Pipeline

# In CI/CD or pre-commit hook
python skills/humanize-ai-text/scripts/detect.py content.md -s
# Fail if AI score > threshold

Troubleshooting

False Positives

Some legitimate writing may trigger patterns. Review manually.

False Negatives

AI models evolve. Update patterns.json regularly.

Encoding Issues

Ensure files are UTF-8 encoded:

python skills/humanize-ai-text/scripts/detect.py file.txt --encoding utf-8

Examples

Example 1: Blog Post

# Check AI score
python skills/humanize-ai-text/scripts/detect.py blog_post.md -s
# Output: AI Probability: 68%

# Transform
python skills/humanize-ai-text/scripts/transform.py blog_post.md -o blog_post_clean.md -a

# Compare
python skills/humanize-ai-text/scripts/compare.py blog_post.md -o blog_post_clean.md

Example 2: Email

echo "I hope this email finds you well. I wanted to reach out regarding..." | \
  python skills/humanize-ai-text/scripts/detect.py

Example 3: Academic Writing

# Detect AI patterns in paper
python skills/humanize-ai-text/scripts/detect.py paper.md -j > report.json

# Transform with conservative settings
python skills/humanize-ai-text/scripts/transform.py paper.md -o paper_clean.md

Resources

• Wikipedia: "Signs of AI writing"
• Original research on AI text detection
• Community pattern contributions

---

For more info, visit: https://clawhub.ai/biostartechnology/humanizer