The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @membranehq/cli package from npm. This is the official command-line interface provided by the vendor (Membrane) to interact with their services.
[COMMAND_EXECUTION]: It utilizes the membrane CLI to perform various tasks such as authentication (membrane login), connection management (membrane connect), and executing specific LaGrowthMachine actions (membrane action run).
[DATA_EXFILTRATION]: No unauthorized data exfiltration was detected. The skill facilitates legitimate data exchange between the user's environment and the LaGrowthMachine API via the Membrane proxy, with authentication handled server-side.
[PROMPT_INJECTION]: The skill functions as an interface for external data (e.g., lead information and messages), which represents a potential surface for indirect prompt injection if the retrieved content contains instructions for the AI. However, the skill is designed for specific automation tasks and does not bypass safety controls.
Ingestion points: Retrieves lead data, campaign statistics, and member lists from LaGrowthMachine (SKILL.md).
Boundary markers: None explicitly mentioned in the skill's instruction set.
Capability inventory: Includes capabilities to send email and LinkedIn messages, update lead statuses, and manage webhooks (SKILL.md).
Sanitization: Execution relies on the Membrane platform's internal handling of API requests and the AI's standard safety layers.

lagrowthmachine