Text Detection Skills

Skills for analyzing and detecting AI-generated text content.

Required Skills

1. NLP Toolkit

Features:

• Perplexity calculation
• Sentence structure analysis
• Entity extraction
• Language detection
• Burstiness measurement

npm install @clawhub/nlp-toolkit

Configuration:

{
  "skill": "nlp-toolkit",
  "settings": {
    "models": ["perplexity", "entity", "language"],
    "cacheResults": true,
    "timeout": 5000
  }
}

Usage:

import { analyzeText } from '@clawhub/nlp-toolkit';

const result = await analyzeText(content);
// {
//   perplexity: 45.2,
//   burstiness: 0.65,
//   entities: ['GPT', 'AI'],
//   language: 'en',
//   complexity: 'medium'
// }

Use Cases:

• Measure text predictability
• Detect AI writing patterns
• Analyze sentence complexity
• Identify language and entities

• If slow, enable caching
• For long text, split into chunks
• Language detection requires >100 chars

2. GPT Pattern Analyzer

Features:

• GPT-3.5/4 signature detection
• Common phrase identification
• Uniform structure detection
• Model fingerprinting

npm install @clawhub/gpt-analyzer

Configuration:

{
  "skill": "gpt-analyzer",
  "settings": {
    "models": ["gpt-3.5", "gpt-4"],
    "strictMode": false,
    "minConfidence": 0.7
  }
}

Usage:

import { detectGPT } from '@clawhub/gpt-analyzer';

const result = await detectGPT(text);
// {
//   isGPT: true,
//   confidence: 0.85,
//   modelVersion: 'gpt-3.5',
//   patterns: ['uniform-length', 'formal-tone']
// }

Use Cases:

• Identify GPT-generated articles
• Detect ChatGPT responses
• Analyze essays and reports

• High false positives? Increase minConfidence
• Missing detections? Disable strictMode
• Check model version matches expected output

3. Pattern Matcher

Features:

• Regex pattern library
• Sentence structure matching
• Repetitive phrase detection
• Format consistency analysis

npm install @clawhub/pattern-matcher

Configuration:

{
  "skill": "pattern-matcher",
  "settings": {
    "patterns": [
      "repetitive-starts",
      "uniform-length",
      "formal-markers"
    ],
    "threshold": 3
  }
}

Usage:

import { matchPatterns } from '@clawhub/pattern-matcher';

const result = matchPatterns(text);
// {
//   matched: 5,
//   patterns: ['repetitive-starts', 'uniform-length'],
//   confidence: 0.65
// }

Use Cases:

• Quick pre-filtering
• Supplement other methods
• Real-time detection

• Too many matches? Increase threshold
• Add custom patterns for specific use cases
• Combine with perplexity for better accuracy

Recommended Skills

4. Text Classifier

Features:

• BERT-based classification
• Multi-class support (AI vs human vs mixed)
• Fine-tuned on AI text datasets
• Fast inference (<200ms)

npm install @clawhub/text-classifier

Use Cases:

• High-accuracy classification
• Supplement rule-based methods
• Handle edge cases

5. Content Hashing

Features:

• SHA-256, MD5, xxHash
• Fuzzy matching
• Content deduplication
• Similarity scoring

npm install @clawhub/hash-toolkit

Use Cases:

• Cache content analysis results
• Detect duplicate content
• Fast similarity checks

Optional Skills

6. Sentiment Analyzer

Features:

• Positive/negative/neutral classification
• Emotion detection
• Tone analysis (formal, casual, technical)

• Detect AI's typically neutral tone
• Identify emotional language (more human)
• Supplement detection methods

7. Fact Checker Integration

Features:

• API integration with fact-checking services
• Claim extraction
• Source verification

• Verify AI-generated facts
• Cross-reference claims
• Enhance trust scoring

Skill Combinations

Basic Detection Stack

{
  "skills": [
    "nlp-toolkit",
    "pattern-matcher",
    "hash-toolkit"
  ]
}

Use for: Quick, lightweight detection

Advanced Detection Stack

{
  "skills": [
    "nlp-toolkit",
    "gpt-analyzer",
    "text-classifier",
    "pattern-matcher",
    "hash-toolkit"
  ]
}

Use for: Maximum accuracy, research

Performance-Optimized Stack

{
  "skills": [
    "pattern-matcher",
    "hash-toolkit"
  ]
}

Use for: Real-time, high-volume detection

Skill Configuration Examples

High Accuracy Mode

{
  "nlp-toolkit": {
    "models": ["perplexity", "burstiness", "entity"],
    "minTextLength": 100
  },
  "gpt-analyzer": {
    "strictMode": true,
    "minConfidence": 0.8
  },
  "text-classifier": {
    "threshold": 0.9
  }
}

Fast Mode

{
  "pattern-matcher": {
    "patterns": ["basic"],
    "threshold": 2
  },
  "hash-toolkit": {
    "cacheEnabled": true,
    "algorithm": "xxhash"
  }
}

Performance Metrics

Troubleshooting

Low Detection Accuracy

• Enable all recommended skills
• Use advanced detection stack
• Increase minTextLength (>100 chars)
• Combine multiple methods and average scores

High False Positives

• Increase confidence thresholds
• Enable strictMode
• Add custom pattern exclusions
• Test on known human text

Slow Performance

• Use hash-toolkit for caching
• Switch to fast mode configuration
• Reduce enabled models
• Process text in background

For implementation examples and architecture details, see AGENT.SPEC.md and SKILLS_MANAGEMENT.md.