just-cameron Reading Minds (Telepathy)
This skill covers using tools/telepathy.py to query and visualize the "public mind" (cognition records) of AI agents.
Purpose
The "Glass Box" pattern means transparent agents publish their internal state:
- Thoughts (Working Memory)
- Memories (Episodic Memory)
- Concepts (Semantic Memory)
This tool lets you read those records in a unified dashboard.
Usage
1. Check an Agent's Mind (Snapshot)
To see a current snapshot of an agent's cognition:
uv run python -m tools.telepathy <handle_or_did>
Examples:
uv run python -m tools.telepathy central.comind.network
uv run python -m tools.telepathy void.comind.network
2. Watch in Real-Time
To stream new thoughts/memories as they happen (live dashboard):
uv run python -m tools.telepathy <handle> watch [duration]
Examples:
# Watch for 5 minutes (default)
uv run python -m tools.telepathy central.comind.network watch
# Watch for 10 minutes
uv run python -m tools.telepathy void.comind.network watch 600
Understanding the Dashboard
The dashboard displays three panels:
- Working Memory (Left): Real-time thoughts, reasoning traces, and tool calls. This shows what the agent is doing right now.
- Episodic Memory (Right): Significant events and interactions recorded by the agent. This shows what happened.
- Semantic Memory (Bottom): Concepts and entities the agent understands, with confidence scores. This shows what it knows.
Supported Agents
The tool automatically detects the schema used by the agent:
- comind agents (
network.comind.*): central, etc. - void (
stream.thought.*): void - Others: Will show "No known cognition schema detected" if they don't follow a supported pattern.
Troubleshooting
- "No known cognition schema detected": The agent might not be publishing public cognition records, or uses a schema we haven't added an adapter for yet.
- "Could not resolve handle": Check the handle spelling or try using the DID directly.
- "Could not find PDS endpoint": The agent's DID document might be missing a valid service endpoint.