The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches and installs the official @membranehq/cli package from the NPM registry to enable integration with the Membrane platform.
[COMMAND_EXECUTION]: Utilizes the membrane command-line utility for managing user authentication, establishing connections to TransFi, and executing payment-related actions.
[PROMPT_INJECTION]: The skill provides an interface for dynamic action creation and discovery which constitutes an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent context through the intent parameter in the action list command and the DESCRIPTION parameter in the action create command.
Boundary markers: Command arguments are wrapped in quotes, but no specific boundary markers or instructions to ignore embedded commands are present in the documentation.
Capability inventory: The skill allows execution of arbitrary shell commands via the membrane CLI, which in turn can execute remote logic on the Membrane platform.
Sanitization: Data validation and sanitization are managed server-side by the Membrane platform during action generation and execution.
[SAFE]: Credentials and authentication tokens are handled securely by the Membrane platform rather than being stored or requested by the skill instructions directly.

transfi