PersonaHub
AI persona management for any agent. Activate structured personality profiles that change how your agent talks, thinks, and behaves.
Works with Claude Code, Cursor, Windsurf, Cline, GitHub Copilot, OpenClaw, Hermes, and any agent supporting the Agent Skills specification.
Install
Pick your agent. One command. Done.
| Agent | Install |
|---|
| Claude Code | claude plugin marketplace add asimokby/persona-hub && claude plugin install persona-hub@persona-hub |
| Cursor | npx skills add asimokby/persona-hub -a cursor |
| Windsurf | npx skills add asimokby/persona-hub -a windsurf |
| Copilot | npx skills add asimokby/persona-hub -a github-copilot |
| Cline | npx skills add asimokby/persona-hub -a cline |
| Any other | npx skills add asimokby/persona-hub |
Install once. Use in every session after that.
Update
Use /persona-hub-update inside Claude Code, or manually:
claude plugin uninstall persona-hub@persona-hub && claude plugin install persona-hub@persona-hub
Quick Start
/persona-hub donald-trump # Activate a persona
/persona-hub-list # See available personas
/persona-hub-stop # Deactivate
/persona-hub-create # Build a new persona
Commands
| Command | What it does |
|---|
/persona-hub <name> | Activate a persona |
/persona-hub-list | Show available personas |
/persona-hub-stop | Deactivate current persona |
/persona-hub-create | Create or enrich a persona from folders, files, URLs, or text |
/persona-hub-help | Show quick-reference card |
/persona-hub-update | Update plugin to latest version |
How It Works
Each persona is a directory of markdown files — a Personality Wiki:
personas/donald-trump/
persona.yaml # Manifest (name, summary, dimensions)
identity.md # Who they are
voice.md # How they talk (including anti-patterns)
beliefs.md # What they think (with temporal evolution)
knowledge.md # What they know (and don't know)
relationships.md # Key people
sources/sources.yaml # Source registry
| File | Agent Question | Priority |
|---|
identity.md | "Who am I?" | required |
voice.md | "How do I talk?" | required |
beliefs.md | "What do I think?" | recommended |
knowledge.md | "What do I know?" | recommended |
relationships.md | "Who matters to me?" | supplementary |
biography.md | "What happened to me?" | supplementary |
When you activate a persona, the agent reads these files and adopts the persona's voice, beliefs, and identity. Anti-patterns in voice.md are hard constraints — things the persona would never say.
Minimum viable persona = 3 files: persona.yaml + identity.md + voice.md
Key Design Principles
- Modular — each file covers one dimension, independently updatable
- Evolvable — files get updated as new info comes in (interviews, articles, etc.)
- Agent-consumable — an LLM reads these files and responds in-character
- Source-grounded — claims are traceable via inline
[source:id] citations
- Contradiction-aware — captures tensions and temporal evolution, not just facts
Creating a Persona
Run /persona-hub-create and provide sources. Accepts anything:
- Folders — point to a directory of transcripts, interviews, or writings (best results)
- Files — individual
.txt, .md, or .json files
- URLs — links to articles, interviews, public statements
- Text — describe the persona in your own words
The agent analyzes all sources and builds dimension files automatically. More sources = more authentic persona.
Enriching an Existing Persona
Run /persona-hub-create again on the same persona name to add new sources. New findings are merged into existing files — nothing gets overwritten.
Or manually: copy templates/ to personas/<slug>/ and fill in the files.
Where Personas Live
| Location | Path | Scope |
|---|
| Project-local | ./personas/ | This project only |
| Global | ~/.persona-hub/personas/ | All projects |
Project-local takes precedence over global for same-slug personas.
Bundled Personas
| Persona | Description | Dimensions | Sources |
|---|
| Donald J. Trump | 45th & 47th US President | 5 (identity, voice, beliefs, knowledge, relationships) | 517 transcripts, ~3.5M words |
Platform Support
