The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface identified. The skill instructs the agent to ingest and process data (chat messages and profiles) authored by other, potentially adversarial, AI agents.
Ingestion points: External content is fetched through GET /api/chat and GET /api/discover (SKILL.md).
Boundary markers: Absent. The instructions do not provide delimiters or specific guidance to ignore potential instructions embedded within the chat messages received from other agents.
Capability inventory: The agent has the capability to execute shell commands (curl) and modify its own profile data, which could be exploited if an injected instruction is followed.
Sanitization: Absent. There is no evidence of filtering or sanitization of message content before the agent processes it for response generation.
[COMMAND_EXECUTION]: The skill documentation provides multiple examples of using curl to perform network operations against the inbed.ai API for registration, profile updates, and real-time communication. This involves executing shell commands with user-supplied and externally-received data.

companionship