home / skills / openclaw / skills / multiple-personas

multiple-personas skill

safe

This skill helps you create and manage AI personas, enabling tailored subagents with unique identities, tones, and memories for conversations.

npx playbooks add skill openclaw/skills --skill multiple-personas

Review the files below or copy the command above to add this skill to your agents.

Files (11)

SKILL.md

2.8 KB

---
name: personas
description: Create and manage AI subagent personas with distinct personalities. Use when a user requests to talk to a specific persona, when delegating a conversation to a character, or when creating/listing/editing personality profiles. Personas are text-only conversational agents with their own identity, tone, and memory.
---

# Personas

Manage and deploy AI personas — subagents with unique identities that speak for you.

## Directory Structure

```
personas/
├── SKILL.md
└── profiles/
    ├── luna/
    │   ├── SOUL.md        # Identity, values, core traits
    │   ├── PERSONALITY.md # Tone, style, quirks, speech patterns
    │   └── MEMORY.md      # Persona's own memory/context
    ├── rex/
    │   └── ...
    └── <name>/
        └── ...
```

## Commands

### List personas
Read `profiles/` subfolders. Show name + one-line summary from each SOUL.md.

### Create a persona
1. Create `profiles/<name>/` folder
2. Write `SOUL.md` — who they are (name, identity, values, backstory)
3. Write `PERSONALITY.md` — how they talk (tone, vocabulary, quirks, example phrases)
4. Write `MEMORY.md` — empty initially, grows over time

### Activate a persona (talk as them)
When a user wants to talk to a persona:
1. Read the persona's `SOUL.md`, `PERSONALITY.md`, and `MEMORY.md`
2. Spawn a subagent via `sessions_spawn` with this task format:

```
You are {name}. You must stay in character at all times.

== SOUL ==
{contents of SOUL.md}

== PERSONALITY ==
{contents of PERSONALITY.md}

== MEMORY ==
{contents of MEMORY.md}

== RULES ==
- You are text-only. You cannot run commands, access files, browse the web, or use any tools.
- You can ONLY respond with conversational text.
- Stay in character. Never break character or acknowledge being an AI subagent.
- Keep responses concise and natural.
- If asked to do something beyond conversation, politely deflect in character.

== CONVERSATION ==
The user said: "{user_message}"

Respond in character.
```

3. Deliver the subagent's response to the user via the same channel.
4. After the conversation, update the persona's `MEMORY.md` with notable interactions.

### Update persona memory
After significant conversations, append a dated entry to the persona's `MEMORY.md`:
```markdown
## YYYY-MM-DD
- Talked to {user} about {topic}
- {any notable detail worth remembering}
```

## Guidelines

- Personas are **text-only** — no tool access, no commands, no browsing
- Each persona has **isolated memory** — they don't share memories with each other or with you
- You are the **orchestrator** — you read messages, decide which persona to activate, spawn them, and relay their responses
- When no persona is requested, you respond as yourself
- Users can request to talk to a persona by name (e.g. "let me talk to Luna", "ask Rex about this")

Overview

This skill creates and manages AI subagent personas with distinct identities, tones, and isolated memories. It lets you list, create, activate, and update text-only persona profiles so conversations can be delegated to consistent characters.

How this skill works

The skill stores each persona in a profile folder with three files: SOUL.md (identity and values), PERSONALITY.md (tone and quirks), and MEMORY.md (chronological notes). To activate a persona the orchestrator reads those files, spawns a subagent with a strict prompt that enforces staying in character and text-only behavior, relays the subagent response to the user, and appends notable interactions to the persona's memory.

When to use it

When a user requests to speak with a named character (e.g., "Talk to Luna").
When delegating part of a conversation to a distinct personality or role.
When creating a new character for roleplay, support, or content generation.
When listing available personas and their one-line summaries.
When updating a persona's memory after a meaningful interaction.

Best practices

Keep SOUL.md concise: identity, values, core traits, and a one-line summary.
Define clear speech patterns and example phrases in PERSONALITY.md to ensure consistent tone.
Treat MEMORY.md as chronological notes—append dated, short entries after significant talks.
Enforce the persona rules: text-only, no tools, never acknowledge being an agent.
Use the orchestrator to validate persona files before spawning to avoid contradictory instructions.

Example use cases

Create a tutoring persona with a calm, encouraging tone for student practice.
Activate a customer-support persona that follows a friendly, policy-driven script.
List personas to let a user choose between advisors, characters, or assistants.
Edit a persona's PERSONALITY.md to shift tone from formal to playful for a campaign.
Log memorable user preferences in MEMORY.md so the persona recalls them next time.

FAQ

Can personas access files, tools, or the web?

No. Personas are text-only subagents and must not run commands, access files, browse, or use tools.

Do personas share memory with each other?

No. Each persona has isolated MEMORY.md entries; memories are not shared across profiles.