The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the official @membranehq/cli tool from the npm registry, which is a vendor-owned resource used for platform interactions.
[COMMAND_EXECUTION]: Employs the membrane CLI to perform operations such as authentication, connection management, and executing platform actions, which involves running shell commands on the system.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it retrieves and processes data from external sources.
Ingestion points: Untrusted data enters the agent's context through membrane action run and membrane request commands when fetching surveys, questions, and responses.
Boundary markers: No explicit delimiters or instructions are provided to help the agent distinguish between integration logic and potentially malicious instructions embedded in external data.
Capability inventory: The agent has the capability to execute platform-specific commands and make network requests via the membrane proxy service.
Sanitization: The instructions do not define any validation or sanitization steps for the data retrieved from external API endpoints before it is utilized in the workflow.

disqo