The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to influence the agent's behavior.\n- Ingestion points: Ingests user messages and document content through the chatStream and streamRAG functions.\n- Boundary markers: The implementation examples lack explicit boundary markers or instructions to treat streaming data as untrusted content.\n- Capability inventory: The agent context includes permissions for Read, Write, Edit, and Bash(curl:*), which could be targeted by successful injection attacks.\n- Sanitization: The provided code demonstrates direct usage of API output without additional sanitization or filtering logic.\n- [EXTERNAL_DOWNLOADS]: References the cohere-ai and express libraries as prerequisites. These are well-known services commonly used for building AI-integrated web applications.

cohere-webhooks-events