The Agent Skills Directory

Indirect Prompt Injection (LOW): Surface for indirect prompt injection through processed message content. Ingestion points: message.content and message.title in SKILL.md and references/examples.md. Boundary markers: Absent in code examples. Capability inventory: No high-risk capabilities like arbitrary command execution or system-level modification were found in the provided scripts. Sanitization: Incoming message content is not sanitized in the provided examples.
Unverifiable Dependencies & Remote Code Execution (LOW): The skill suggests installing rns and lxmf packages via pip. While these are the official and expected libraries for the protocol described, the author does not belong to the predefined list of trusted organizations.

lxmf-protocol