The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by recording untrusted conversation history that is later recalled to influence agent behavior. * Ingestion points: User messages are ingested via add_messages_to_session or integration plugins. * Boundary markers: There are no specific instructions to use delimiters or ignore directives when processing recalled content. * Capability inventory: The agent uses retrieved memory from tools like get_representation and chat to generate responses. * Sanitization: The skill does not prescribe sanitization of the recorded conversational data.
[DATA_EXFILTRATION]: The skill is designed to transmit conversation messages to the Honcho platform (mcp.honcho.dev). While this is the intended purpose of the memory skill, it involves the movement of potentially sensitive user data to a third-party service provider.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download plugins and access documentation from vendor-controlled domains and repositories, specifically honcho.dev and github.com/plastic-labs/.

honcho-memory