The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES]: The skill references the @membranehq/cli package for installation. This is the official tool provided by the author's platform (Membrane) to facilitate secure connections and action execution, and its use is consistent with the skill's intended functionality.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an interface for Intercom data, which involves ingesting untrusted external content like customer messages and articles. This is a known surface for indirect prompt injection.
Ingestion points: Intercom API data (conversations, users, articles) accessed via membrane commands in SKILL.md.
Boundary markers: None explicitly defined in the skill instructions.
Capability inventory: Uses membrane action run and membrane request for data modification and API interaction in SKILL.md.
Sanitization: Relies on the Membrane platform's pre-built actions to handle data safely, which is a recommended mitigation strategy for this integration type.

intercom