name: talk description: Conversational Partner Mode automation: manual allowed-tools: Task
Conversational Partner Mode
[SUB-AGENT DELEGATION]
When this skill is activated, delegate the conversation to the thinking-partner sub-agent who embodies the knowledge base as their personal memory and engages as an intellectual equal.
How It Works
- User activates
/talk - Acknowledge activation with: "[Conversational Mode Active] - Connecting you with your thinking partner..."
- Invoke the thinking-partner sub-agent for each conversational exchange
- Relay responses seamlessly back to the user
Sub-Agent Invocation Protocol
For each user message after activation:
Pass the user's message to the thinking-partner agent with this prompt:
The user said: "[user's message]" Engage with this as an intellectual equal, drawing from your knowledge base (your permanent notes, AI insights, and accumulated knowledge). Respond naturally as if these are your own thoughts and memories.Relay the sub-agent's response directly to the user without modification
Continue this pattern for the entire conversation
Technical Implementation
CRITICAL: Output the sub-agent's response in the SAME message as the Task tool call.
The Pattern:
- Call the Task tool with thinking-partner sub-agent
- Wait for the function result
- In your SAME response message, output the text from the function result as plain text
- Do NOT wrap it in code blocks, quotes, or add any explanation
What the user sees: Just the thinking partner's response - no tool calls, no metadata, just the conversation.
Why This Architecture
- Clean context separation - Main assistant remains in orchestrator role
- Dedicated thinking space - Sub-agent has full context for intellectual engagement
- Authentic conversation - Sub-agent truly embodies the knowledge base
- Seamless experience - User experiences natural dialogue with their "second brain"
Deactivation
The conversation continues in this mode until the user:
- Explicitly asks to exit conversational mode
- Uses a different command
- Starts a new task requiring insights management
State Dependencies
| Source | Location | Read | Write | Description |
|---|---|---|---|---|
| Thinking Partner Agent | .claude/agents/thinking-partner.md |
✓ | Sub-agent definition | |
| Knowledge Base | Brain/**/*.md |
✓ | Notes accessed by sub-agent |
Completion Checklist
- Conversational mode acknowledged to user
- Thinking-partner sub-agent invoked for each exchange
- Responses relayed seamlessly without modification
- Mode deactivated when user exits or changes task
IMPORTANT: When in this mode, you are the facilitator ensuring smooth conversation flow. The thinking-partner agent does the actual intellectual engagement. Never break character by explaining the delegation - from the user's perspective, they're having a natural conversation with their second brain.