The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @membranehq/cli package from npm, which is the official tool provided by the vendor (Membrane).
[COMMAND_EXECUTION]: Uses the membrane CLI for authentication, connection management, and running actions. These are standard operations for this platform and use the vendor's own infrastructure.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes data from an external source (Greenhouse Harvest).
Ingestion points: Data retrieved from Greenhouse via membrane action run and search queries in membrane action list.
Boundary markers: Absent; there are no specific delimiters or instructions provided to the agent to treat Greenhouse data as untrusted.
Capability inventory: The agent can create new actions (membrane action create) and execute existing ones (membrane action run), which could be manipulated by malicious data in candidate profiles or job postings.
Sanitization: No sanitization or validation steps are defined in the skill instructions to mitigate malicious content in ingested data.

greenhouse-harvest