The Agent Skills Directory

[SAFE]: The skill defines rules for Matrix @mentions, message routing, and loop prevention. These are benign instructional guidelines for agent behavior.
[INDIRECT_PROMPT_INJECTION]: The skill describes handling task assignment messages and historical context from external sources. It provides instructions to differentiate between historical context and current tasks, which acts as a structural boundary to prevent instruction confusion from ingested data. Ingestion points: Task assignment messages and history sections. Boundary markers: Instructions to treat history as context only. Capability inventory: Tool access mentioned include a message tool and CLI tools organization and hiclaw. Sanitization: The agent is instructed to filter Matrix IDs and assess content significance before sending messages.

communication