The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install the Membrane CLI (@membranehq/cli) globally via NPM. This is a legitimate vendor resource used for interacting with the Membrane platform.
[COMMAND_EXECUTION]: Utilizes the membrane CLI to perform operations such as authentication, connection management, and running CloudTables actions.
[DATA_EXFILTRATION]: Employs a secure browser-based login flow via the CLI (membrane login), ensuring that sensitive credentials like API keys are managed by the vendor's platform and never directly handled or stored by the agent skill.
[PROMPT_INJECTION]: As the skill fetches and processes data from external CloudTables datasets, it possesses a surface for indirect prompt injection. However, the use of structured schemas and predefined actions helps mitigate the risk of instructions being interpreted as commands.
Ingestion points: Data retrieved from CloudTables datasets and action schemas via membrane action run and membrane action list commands.
Boundary markers: Not present in the provided instructions.
Capability inventory: Shell command execution for managing platform resources and running actions via the CLI.
Sanitization: Relies on the platform's internal structured data handling.

cloudtables