The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @membranehq/cli package globally via npm. This is a vendor-owned tool from membranedev used to facilitate authentication and API communication.
[COMMAND_EXECUTION]: The skill uses the membrane command-line interface to perform several operations, including user authentication (membrane login), connecting to Braze (membrane connect), and executing actions or proxy requests (membrane action run, membrane request). These are legitimate vendor functionalities.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it retrieves and processes data from Braze (such as user profiles, campaigns, and email templates) which could contain adversarial instructions intended to influence the agent's behavior. Ingestion points: Data retrieved from the Braze API via actions like export-user-by-id, get-email-template, and list-campaigns. Boundary markers: No specific delimiters or instructions are defined to isolate external data from the agent's instruction context. Capability inventory: The skill can perform actions with side effects, such as send-messages, update-subscription-status, and delete-users using the membrane CLI. Sanitization: No explicit sanitization or validation of the content retrieved from Braze is implemented before it is presented to the agent.

braze