The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the Membrane CLI tool (@membranehq/cli) via npm. This is a vendor-provided tool necessary for the skill's primary function.
[COMMAND_EXECUTION]: The skill relies on executing shell commands through the membrane CLI to manage connections, discover actions, and execute workflows. These commands are scoped to the authenticated user's Membrane account.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes external data from Banked and natural language intents to search for or create new actions.
Ingestion points: Untrusted data enters the context via the output of membrane action list and membrane action run (SKILL.md).
Boundary markers: None identified in the prompt instructions to delimit tool outputs.
Capability inventory: The skill can execute actions (membrane action run) which may include financial transactions and create new executable logic (membrane action create).
Sanitization: No specific sanitization or validation of the remote action schemas or outputs is performed within the skill instructions.

banked