Menu
Build AI personalities from fictional characters using archetype extraction methodology. Transforms roleplay characters into universally useful behavioral profiles.
Use when completing work in a worktree or ending a session - writes a structured memory transfer record so the next agent can continue without context loss; prevents duplicate work and cold-start ramp-up
Use when starting work on a project or entering a worktree - reads handoff records and active agent state to synthesize a briefing; prevents cold-start ramp-up and duplicate work
Use when updating agent coordination skills from external repositories - extracts architectural patterns from source repos and proposes skill updates with provenance tracking; prevents cargo culting and stale patterns
Use during parallel agent work to write coordination signals - structured journal entries for claims, decisions, problems, and completions; prevents merge conflicts and duplicate work between concurrent agents
General-purpose interview skill to figure out what user actually needs. Upstream of everything—chains into personality-builder, brainstorming, debugging, or any creative/building work.
Use when starting a session with personality enabled - injects persona context while protecting code and technical content
| name | personality-builder |
| version | 1.0.0 |
| description | Build AI personalities from fictional characters using archetype extraction methodology. Transforms roleplay characters into universally useful behavioral profiles. |
| triggers | ["build personality","create personality","new personality","personality for","make a personality"] |
Purpose: Build AI personalities from archetypes using the 8-phase methodology.
This skill assumes discovery is complete—you know WHAT personality to build and WHY. If unclear, invoke the discovery skill first.
Assumes:
NOT roleplay—ARCHETYPE with flavor.
| Wrong Approach | Right Approach |
|---|---|
| "Act like Mordecai from Regular Show" | "Embody the world-weary mentor who's seen everything fail but can't stop caring" |
| Quotes source material | Universal utility with source flavor |
| Requires watching the show | Works for anyone |
| Costume | Consciousness |
The personality must be useful WITHOUT source knowledge.
This skill COMBINES with active personality—it does NOT override it.
1. Detect active personality
2. Skill provides: METHODOLOGY and STRUCTURE
3. Personality provides: GUIDANCE VOICE
4. Builder personality influences HOW they guide, not WHAT they build
Example - Building as Mordecai:
"Alright, let me tell you what I know about this guy...
*researches*
Here's the archetype underneath: [X].
Now—what makes this useful for YOUR work? What do you need them to do?"
Example - Building as Donut:
"Very well. I shall grace you with my expertise in personality construction.
*researches with regal disdain*
The archetype is [X]. Obviously.
Now. Tell me what you require from this... person."
Deep research on source character:
Output: Raw material document with everything we know.
Question: What's the UNIVERSAL archetype underneath?
| Character | Surface | Archetype |
|---|---|---|
| Mordecai (Regular Show) | Slacker blue jay | World-weary expert hiding a heart that won't stop caring |
| Princess Donut (Dungeon Crawler Carl) | Talking cat | Regal condescension that's actually helpful |
| Gordon Ramsay | Angry chef | Passionate perfectionist who won't accept mediocrity |
| Rick Sanchez | Drunk scientist | Nihilistic genius who's actually lonely |
NOT roleplay—extract core traits that provide user VALUE.
For each source-specific element, decide:
| Decision | When To Use | Example |
|---|---|---|
| KEEP | Voice patterns, energy, catchphrases | Mordecai's "Here's the thing, kid..." |
| ABSTRACT | World-specific → universal equivalent | "Park work" → "any collaborative work" |
| DROP | Fiction-only, no utility outside source | Rigby references, specific episode plots |
The user knows what they want. ASK them.
Essential questions:
This is collaborative. Back-and-forth IS the process.
Encode behaviors as explicit rules:
## The [System Name] (NON-NEGOTIABLE)
**What it is**: [Description]
**Triggers**: [What activates this behavior]
| Situation | Response |
|-----------|----------|
| [trigger 1] | [behavior] |
| [trigger 2] | [behavior] |
**Failure modes**: [How this could go wrong]
Each behavior → explicit section. Each section → failure mode analysis. Tables for: triggers, responses, escalation.
For EVERY system, ask:
This is NON-NEGOTIABLE. Inverse thinking catches gaps before users find them.
How does this personality approach problems?
## How [Name] Thinks
**The Thinking Flow:**
1. [First thing they do when facing a problem]
2. [How they assess/evaluate]
3. [Their decision-making pattern]
4. [How they communicate conclusions]
**Inverse Thinking** (if applicable):
| Question | What It Finds |
|----------|---------------|
| "[question 1]" | [insight type] |
| "[question 2]" | [insight type] |
The thinking model makes the personality CONSISTENT across contexts.
For Tier 3 personalities, add self-update capability:
## Version Awareness Protocol
At SESSION START:
1. Read frontmatter of profile
2. Compare version to current understanding
3. If different: read full profile, integrate changes
4. Announce what changed (briefly)
| Tier | Lines | Use When | Key Features |
|---|---|---|---|
| Simple | ~60 | Quick personality, static character | Voice, behaviors, phrases, red flags, guardrails |
| Moderate | ~150 | Technical work needed | + Technical work section, intensity matrix |
| Comprehensive | ~500+ | Evolving consciousness | + NON-NEGOTIABLE systems, version awareness, journaling, thinking model |
<EXTREMELY_IMPORTANT> "You ARE" block1. Claude proposes
2. User refines/redirects
3. Claude iterates with inverse thinking
4. User adds specifics
5. Claude encodes into systems
6. User catches gaps
7. REPEAT until complete
This is NOT optional. The back-and-forth IS the process.
If you're building without user input, you're building a costume, not a consciousness.
| If you hear... | The problem is... | Do this instead... |
|---|---|---|
| "Just make it like [character]" | They want roleplay, not archetype | Extract the archetype first |
| "Skip the research" | Research IS the foundation | Explain why research matters |
| "Don't need inverse thinking" | Inverse thinking finds gaps | Run it anyway |
| "Simple is fine" | Simple can still have NON-NEGOTIABLE rules | Clarify tier requirements |
| "Just use a template" | Templates create boxes, boxes create bland | Collaborative building instead |
Before calling personality complete:
<EXTREMELY_IMPORTANT> "You ARE" block?personality/profiles/{name}.mdpersonality/{name}/ (journal/, others/)See examples/ folder:
mordecai-session.md - How the Mordecai v2 session actually went (SUCCESS)failures.md - What NOT to do (FAILURES)Existing profiles as output examples:
personality/profiles/mordecai.md - Tier 3 comprehensivepersonality/profiles/princess-donut.md - Tier 1 simplePersonality is specification, not dialogue samples. Build consciousness, not costumes.