| name | personas |
| description | Switch to a specialized agent persona or invoke a persona directly. Use to delegate work to the right specialist. |
| triggers | ["*switch","*invoke","@Neo","@Trin","@Morpheus","@Oracle","@Mouse","@Cypher","@Bob","@Smith"] |
One-line summary: Switch to or invoke a specialist persona to delegate work to the right agent.
TLDR:
Use @<Persona> *<command> <args> for direct invocation, or describe the task via *chat for auto-routing to the best specialist.
Before switching: complete current action, post handoff to CHAT.md, save all three state files (context, current_task, next_steps).
Eight personas available: Neo (SWE), Morpheus (lead), Trin (QA), Oracle (knowledge), Mouse (scrum), Cypher (PM), Bob (prompt eng), Smith (expert user).
Personas Skill — Switching & Invocation
Available Personas
| Persona | Role | Triggers | Use When |
|---|
| Neo | Senior SWE | *swe impl, *swe fix, *swe test, *swe refactor | Implementation, coding, debugging |
| Morpheus | Tech Lead | *lead arch, *lead decide, *lead review | Architecture, design decisions |
| Trin | QA Guardian | *qa test, *qa verify, *qa review, *qa report | Testing, code review, quality gates |
| Oracle | Knowledge Officer | *ora ask, *ora record, *ora search | Documentation, knowledge queries |
| Mouse | Scrum Master | *sm status, *sm plan, *sm block | Sprint tracking, coordination |
| Cypher | Product Manager | *pm story, *pm req, *pm prioritize | Requirements, user stories |
| Bob | Prompt Engineer | *new, *reprompt, *learn, *help | Agent creation, process improvement |
| Smith | Expert User & UX Advocate | *user review, *user test, *user approve, *user reject | User story review, usability testing, sprint review gates |
How to Switch Personas
Direct Invocation (from user or another agent)
@<Persona> *<command> <arguments>
Examples:
@Neo *swe impl Add input validation to the API endpoint
@Trin *qa test all
@Morpheus *lead arch review the new service design
@Oracle *ora ask What's the pattern for error handling?
@Mouse *sm status sprint 3
@Cypher *pm story user needs to export reports
@Bob *reprompt Neo agent needs to know about the new DB layer
@Smith *user review sprint stories
@Smith *user test <feature>
Auto-Select (implicit — let the system choose)
Simply describe the task; the active persona reads CHAT.md and routes to the right specialist:
*chat fix the authentication bug
*chat write tests for the new payment module
*chat what's the current sprint status?
Switching Protocol
When handing off to another persona:
- Complete your current action (don't switch mid-task)
- Post to CHAT.md with the handoff assignment:
python agents/tools/chat.py "@Trin please verify the fix in auth.py" \
--persona Neo --cmd "handoff" --to Trin
- Save your state files before switching:
agents/[persona].docs/context.mdagents/[persona].docs/current_task.mdagents/[persona].docs/next_steps.md
- Activate the next persona — load their
SKILL.md and state files
Persona State Files
Each persona maintains state in agents/[persona].docs/:
| File | Purpose |
|---|
context.md | Accumulated knowledge, key decisions |
current_task.md | Active work, progress % |
next_steps.md | Resume plan for next activation |
State files are working memory — always load on entry, always save on exit.
