Personality DNA -- Compact Portable Identity

Distills the entire personality model into a single ~2000-token JSON document that can be versioned, exported, and used to cold-start a new instance.

How It Works

1. Compile -- Read all personality data (user model, style profiles, exemplar fingerprints)
2. Compress -- Distill into the most essential patterns, values, and style markers
3. Export -- Save as a portable JSON file at ~/.nomos/personality-dna.json
4. Import -- Load DNA to seed an empty user model (cold-start a new instance)

Backend tool (use this, don't improvise)

Call the `personality_dna` tool (mcp__nomos-think__personality_dna) to compile the DNA -- it runs the real compiler (top decision patterns + values + style genome + exemplar fingerprints, within the ~2000-token budget) and returns a preview. For /dna export / /dna show, pass export: true to get the full portable document. Do not hand-assemble the DNA; use the compiler's output.

Commands

• /dna export -- Compile and export personality DNA
• /dna import or /dna import <file> -- Import DNA from file
• /dna show -- Preview the current DNA without exporting
• /dna diff -- Compare current personality against last exported DNA

Export Protocol

When the user invokes /dna export:

Phase 1: Gather

1. Call user_model_recall to load the full user model
2. Call memory_search for exemplars
3. Load style profile data if available

Phase 2: Compile

Build the DNA document with these sections:

{
  "version": "1.0",
  "compiled_at": "ISO timestamp",
  "identity": {
    "summary": "2-3 sentence description of who this person is",
    "roles": ["role1", "role2"],
    "expertise": ["area1", "area2"]
  },
  "decision_patterns": [
    {
      "principle": "...",
      "context": "...",
      "weight": 0.9,
      "exceptions": ["..."]
    }
  ],
  "values": [
    {
      "value": "...",
      "description": "...",
      "rank": 1
    }
  ],
  "style_genome": {
    "formality": 3,
    "tone": "direct",
    "avg_length": "moderate",
    "emoji_usage": "rare",
    "vocabulary_markers": ["word1", "word2"],
    "punctuation_style": "standard",
    "signature_phrases": ["phrase1"]
  },
  "behavioral_signatures": {
    "response_speed": "quick",
    "detail_preference": "moderate",
    "question_style": "direct",
    "conflict_approach": "collaborative"
  },
  "exemplar_fingerprints": [
    {
      "text": "short representative message",
      "context": "slack_work"
    }
  ]
}

Phase 3: Review

Present the compiled DNA to the user:

Personality DNA compiled successfully.

Identity: [summary]
Decision patterns: [count] (top 3 listed)
Values: [count] (top 3 listed)
Style markers: [key characteristics]
Exemplar fingerprints: [count]

Total size: ~XXXX tokens

Export to ~/.nomos/personality-dna.json?

Phase 4: Export

After user confirmation:

1. Write to ~/.nomos/personality-dna.json
2. Also store in config DB under key personality.dna
3. Report success

Import Protocol

When the user invokes /dna import:

Phase 1: Load

1. Read from the specified file (default: ~/.nomos/personality-dna.json)
2. Validate the DNA structure

Phase 2: Preview

Show what will be imported:

• Number of patterns, values, style markers
• Whether it will merge with or replace existing data

Phase 3: Inflate

For each section of the DNA:

1. Decision patterns -- create user_model entries with confidence 0.7 (imported, not directly observed)
2. Values -- create user_model entries with confidence 0.7
3. Style genome -- create/update global style profile
4. Exemplar fingerprints -- store as exemplar memory chunks

Phase 4: Report

DNA imported successfully.
Created: X decision patterns, Y values, Z style markers
Starting confidence: 0.7 (will increase as patterns are confirmed through interaction)

Important Rules

• Compress ruthlessly -- the DNA must be under ~2000 tokens. Prioritize signal density.
• Top 10 patterns only -- rank by weight, take the most impactful
• Top 5 values only -- rank by confidence
• 3-5 exemplar fingerprints -- the most representative short messages
• Import at 0.7 confidence -- imported data is plausible but unverified
• Never overwrite -- import merges, doesn't replace existing high-confidence entries
• Version the DNA -- include compilation timestamp and version number
• Diff support -- /dna diff shows what changed since last export