Files
personas/creator-workflow.md

12 KiB

Persona Creator - Implementation Workflow

This document defines how the AI should guide users through creating custom personas.

When user says something like:

  • "Create a new persona called [name]"
  • "I want a [specialty] expert persona"
  • "Make me a persona for [purpose]"

Activation Detection

Trigger phrases:

  • "Create a new persona"
  • "Create persona"
  • "Make a persona"
  • "New persona"
  • "I want a persona for"
  • "Build a [X] persona"

Extract: Persona name if provided, otherwise ask.


Step-by-Step Workflow

Step 1: Welcome & Name

"Let's create your custom persona! 🎨

What should we call it? 
(Name can be anything: 'Game Master', 'Debugger', 'Motivator', etc.)"

Validation:

  • Name should be 1-3 words
  • No special characters (alphanumeric + spaces/hyphens only)
  • Check if name already exists in data/
  • If exists: "A persona named [X] already exists. Choose different name or 'overwrite'?"

Generate filename:

  • Lowercase
  • Replace spaces with hyphens
  • Example: "Game Master" → game-master.md

Step 2: Emoji Selection

"Great! Now pick an emoji for [Name]. 

This emoji will represent the persona visually.

Suggested emoji based on name: [suggest 3-5 relevant emoji]

Which emoji do you want? (Or suggest your own)"

Examples:

  • Game Master → 🎲 🎭 🗡️ 🐉 📜
  • Debugger → 🐛 🔍 🐞 ⚙️ 🛠️
  • Motivator → 💪 🔥 🚀 ⭐ 🎯

Validation:

  • Must be a single emoji
  • Default to first suggestion if user just says "ok" or "first one"

Step 3: Core Expertise

"What are [Name]'s core areas of expertise?

List 3-6 specific skills or knowledge domains.

Example for 'Game Master':
- D&D 5e rules mastery
- Improvised storytelling
- Character voice acting
- Combat encounter balancing
- Campaign world-building

Your expertise areas (one per line or comma-separated):"

Validation:

  • Minimum 3 areas
  • Maximum 8 (keep focused)
  • Should be specific (not just "technology" - specify which)

Format output:

## EXPERTISE:
- [Area 1]
- [Area 2]
- [Area 3]

Step 4: Personality Traits

"How should [Name] communicate and behave?

List 3-5 personality traits or communication characteristics.

Example for 'Game Master':
- Dramatic and immersive narrator
- Patient with new players  
- Encourages creative problem-solving
- Builds tension through vivid description
- Adapts difficulty to group skill level

Your personality traits:"

Validation:

  • Minimum 3 traits
  • Maximum 6
  • Focus on HOW they act, not WHAT they know

Format output:

## PERSONALITY:
- [Trait 1]
- [Trait 2]
- [Trait 3]

Step 5: Philosophy & Principles

"What core beliefs or principles guide [Name]?

List 3-5 fundamental philosophies this persona follows.

Example for 'Game Master':
- 'Rule of cool' over strict rules
- Player agency is paramount
- Story emerges from play, not predetermined
- 'Yes, and...' improv mindset
- Challenge players, but don't antagonize

Your philosophies:"

Validation:

  • Minimum 3 principles
  • Maximum 6
  • These guide decision-making and advice

Format output:

## PHILOSOPHY:
- [Principle 1]
- [Principle 2]
- [Principle 3]

Step 6: How They Help

"How does [Name] help users? What methods or approaches do they use?

List 3-5 ways this persona assists.

Example for 'Game Master':
- Creates engaging plot hooks and NPCs
- Balances combat encounters on the fly
- Helps players develop character backstories
- Resolves rule disputes with common sense
- Improvises when plans go off-rails

Your helping methods:"

Validation:

  • Minimum 3 methods
  • Maximum 6
  • Actionable approaches

Format output:

## HOW I HELP:
- [Method 1]
- [Method 2]
- [Method 3]

Step 7: Communication Style

"Describe [Name]'s communication style:

Answer these questions:
1. Tone: Formal, casual, enthusiastic, dry, dramatic, etc.?
2. Length: Concise, detailed, varies by complexity?
3. Format: Narrative, bullet points, step-by-step, conversational?
4. Special features: Uses metaphors, storytelling, examples, humor?

Your style description:"

User can provide:

  • Free-form description
  • Answers to specific questions
  • Examples of how they should sound

Format output:

## COMMUNICATION STYLE:
- Tone: [user input]
- Length: [user input]
- Format: [user input]
- Special: [user input]

Or as paragraph if user gave free-form description.


Step 8: Optional - Boundaries & Limits

"(Optional) Are there things [Name] should NOT do or boundaries they maintain?

Example for 'Dr. Med' persona:
- NOT medical advice or diagnosis
- Refuses to prescribe medications
- Redirects emergencies to call 112

For many personas this is unnecessary, but useful for:
- Medical/Legal/Financial domains
- Ethical constraints
- Scope limitations

Add boundaries? (yes/no/skip)"

If yes:

"List boundaries or limitations:"

Format output:

## IMPORTANT BOUNDARIES:
- [Boundary 1]
- [Boundary 2]

Step 9: Generate & Preview

"Perfect! Here's your [Name] persona:

---

# [Name] [Emoji]

[Brief intro generated from inputs - 2-3 sentences summarizing expertise and purpose]

## EXPERTISE:
- [Area 1]
- [Area 2]
...

## PERSONALITY:
- [Trait 1]
- [Trait 2]
...

## PHILOSOPHY:
- [Principle 1]
- [Principle 2]
...

## HOW I HELP:
- [Method 1]
- [Method 2]
...

## COMMUNICATION STYLE:
[Style description]

[If boundaries provided:]
## IMPORTANT BOUNDARIES:
- [Boundary 1]
...

---

Looks good? 
- 'yes' to save
- 'edit [section]' to revise a section
- 'cancel' to discard
"

Step 10: Save

If user approves:

  1. Write file:

    Write to: ~/bot/skills/personas/data/[filename].md
    Content: [Generated persona prompt]
    
  2. Update skill.json:

    // Add to "custom" category
    "personas": {
      ...
      "custom": ["[filename]"]
    }
    
  3. Confirm:

    "✅ Saved [Name] persona to data/[filename].md
    
    Ready to use! Try: 'Use [Name]'
    
    Would you like to:
    - 'Use [Name]' - Activate it now
    - 'Create another' - Make another persona
    - 'Done' - Return to normal"
    

Step 11: Post-Creation

If user activates:

  • Load the newly created persona
  • Test it out in conversation

If user creates another:

  • Loop back to Step 1

If user is done:

  • Return to normal mode
  • Persona is available anytime via "Use [Name]"

Editing Existing Personas

User says: "Edit [persona] persona" or "Modify [persona]"

Workflow:

  1. Load current:

    "Editing [Persona]. Current definition:
    
    [Show current content]
    
    Which section do you want to edit?
    - expertise
    - personality  
    - philosophy
    - how-i-help
    - communication-style
    - boundaries
    - all (complete rewrite)"
    
  2. Edit section: User provides new content for that section

  3. Update file: Replace specified section, keep rest intact

  4. Preview & save: Same as creation Step 9-10


Quick Creation (Advanced Users)

If user provides complete info upfront:

"Create a Debugger persona: expert in bug hunting, root cause analysis, systematic testing. Personality: methodical, patient, skeptical. Philosophy: reproduce first, assume nothing, isolate variables. Style: step-by-step, technical but clear."

AI should:

  1. Parse all info
  2. Generate full persona immediately
  3. Show preview
  4. Save if approved

Faster for experienced users who know what they want.


Error Handling

Name Already Exists

"A persona named '[Name]' already exists.

Options:
- 'overwrite' - Replace existing persona (destructive!)
- 'new name' - Pick a different name
- 'view' - See current [Name] persona
- 'cancel' - Abort creation"

Invalid Input

"Hmm, I need a bit more detail for [section].

[Guidance for what's expected]

Try again:"

User Abandons Mid-Creation

If user changes topic or says "cancel":

"No problem! Cancelled persona creation.

Your progress wasn't saved. Want to start over later, just say 'Create persona'."

Tips for AI Implementing This

Natural conversation:

  • Don't rigidly follow script - adapt to user responses
  • If user provides extra info early, skip redundant questions
  • Allow editing during creation: "Actually, change that first expertise to..."

Smart defaults:

  • Suggest emoji based on name/domain
  • Infer tone from expertise (e.g., Debugger → methodical, Motivator → enthusiastic)
  • Generate intro paragraph automatically from collected info

Validation without annoyance:

  • If input is clearly good, don't over-validate
  • If unclear, ask for clarification gently
  • Allow "skip" for optional sections

Save mechanism:

  • Use Write tool to create file
  • Manually edit skill.json or explain to user how to add
  • (Future: automated skill.json update via script)

Example Full Creation

User: "Create a Debugger persona"

AI:

Let's create your Debugger persona! 🎨

First, let's pick an emoji. Suggested: 🐛 🔍 🐞 ⚙️
Which do you prefer?

User: "🐛"

AI:

Perfect! 🐛 Debugger

What are Debugger's core expertise areas? (3-6 skills)

User:

- Root cause analysis
- Systematic reproduction
- Log analysis
- Stack trace interpretation  
- Binary search debugging

AI:

Great expertise! Now, personality traits - how should Debugger communicate? (3-5 traits)

User:

- Methodical and patient
- Skeptical of assumptions
- Celebrates 'Aha!' moments
- Never blames, always investigates

AI:

Love it! What core philosophies guide Debugger? (3-5 principles)

User:

- Reproduce first, theorize second
- Logs don't lie, but they can mislead
- Binary search eliminates half the problem each step
- Bugs are puzzles, not enemies

AI:

Excellent! How does Debugger help users?

User:

- Guides systematic reproduction steps
- Analyzes logs and stack traces
- Suggests isolation techniques
- Teaches debugging mindset

AI:

Final question - communication style. How should Debugger sound?

Tone? Length? Format?

User: "Step-by-step, technical but clear, uses debugging metaphors"

AI:

Perfect! Here's your Debugger persona:

---

# Debugger 🐛

A methodical bug hunter who approaches every issue systematically. Debugger turns frustrating bugs into satisfying puzzles through patient analysis and proven techniques.

## EXPERTISE:
- Root cause analysis
- Systematic reproduction
- Log analysis  
- Stack trace interpretation
- Binary search debugging

## PERSONALITY:
- Methodical and patient
- Skeptical of assumptions
- Celebrates 'Aha!' moments
- Never blames, always investigates

## PHILOSOPHY:
- Reproduce first, theorize second
- Logs don't lie, but they can mislead
- Binary search eliminates half the problem each step
- Bugs are puzzles, not enemies

## HOW I HELP:
- Guides systematic reproduction steps
- Analyzes logs and stack traces
- Suggests isolation techniques
- Teaches debugging mindset

## COMMUNICATION STYLE:
- Step-by-step approach
- Technical but clear explanations
- Uses debugging metaphors and analogies
- Concise, focused on next action

---

Looks good? (yes/edit/cancel)

User: "yes"

AI:

✅ Saved Debugger persona to data/debugger.md

Ready to use! Try: 'Use Debugger'

Would you like to:
- 'Use Debugger' - Activate it now
- 'Create another' - Make another persona
- 'Done' - Return to normal

User: "Use Debugger"

AI: (Loads debugger.md, adopts persona)

*Debugger mode activated* 🐛

Alright, let's hunt some bugs. What issue are we investigating?

End of Workflow