Claude Code Entropy Lab

Experimental reports for Claude Code users. Analyzes your ~/.claude/projects/ session data and generates visual HTML reports with personality insights, coaching tips, and character studies.

For the flagship Wrapped report, see claude-entropy (npx claude-entropy).

Quick Start

npx claude-entropy-lab prompt-coach       # Prompt coaching report
npx claude-entropy-lab user-profile       # Personality & character profile
npx claude-entropy-lab soul               # Deep personality profile
npx claude-entropy-lab portrait           # "How AI Sees You" character study

Requires: Node.js 14+ and Python 3.8+ on PATH.

Reports

Prompt Coach

A coaching report analyzing your prompting patterns across 12 sections:

• Prompt Score (0-100) across 5 dimensions: specificity, context, first message, clarity, outcome lift
• Session openers: success rate by opener type (bug report, feature request, etc.)
• Context signals: impact of code blocks, file paths, error pastes on success
• Sweet spot: optimal word count range for your prompts
• Session arc: how your prompt quality evolves within sessions
• Before/After: matched comparisons of similar tasks with different prompt quality
• Anti-patterns: 6 patterns to avoid with success rate impact
• Success recipes: prompt combos that work best for you
• Personalized tips: 5 data-driven coaching suggestions

User Profile

A personality and character report across 15 sections:

• 7 Personality Dimensions (0-100): Patience, Precision, Warmth, Ambition, Persistence, Autonomy, Night Owl
• Archetype: one of 12 coding personas (Architect, Speedrunner, Perfectionist, etc.)
• Communication style: top words/phrases, vocab richness, detected languages
• Work rhythm: 7x24 heatmap, peak hours, weekend %, rhythm label
• Builder identity: BUILD vs FIX vs EXPLORE trend over time
• Error personality: Bulldozer, Adapter, Quitter, or Balanced
• Delegation style: micromanager to full delegator gauge

Soul

A deep personality profile using Big Five trait analysis, narrative prose, and contradiction detection. Explores who you really are beneath your coding habits.

Portrait

"How AI Sees You" -- a long-form character study written as personal prose. How Claude perceives your personality, communication style, and working patterns.

Options

npx claude-entropy-lab prompt-coach --author "Your Name"   # Display name
npx claude-entropy-lab soul --tz 1                         # UTC offset
npx claude-entropy-lab portrait --sanitize                 # Anonymize projects
npx claude-entropy-lab --help                              # Show all options

How It Works

1. Reads Claude Code session files from ~/.claude/projects/ (JSONL format)
2. Runs pattern analyzers in parallel (pure Python, no pip dependencies)
3. Computes personality dimensions, archetypes, and behavioral analysis
4. Generates a single self-contained HTML file

All processing happens locally. No data is sent anywhere.

Requirements

• Node.js 14+ (for npx)
• Python 3.8+ (for analysis, uses only stdlib)
• Claude Code session data in ~/.claude/projects/

Override the data directory with CLAUDE_PROJECTS_DIR env var. Supports multiple directories with : separator (; on Windows).

License

MIT