The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes membrane CLI commands to manage authentication, list connections, and run GraphQL actions.
[EXTERNAL_DOWNLOADS]: Installs the @membranehq/cli package from the official NPM registry to provide the necessary integration tools.
[REMOTE_CODE_EXECUTION]: Utilizes the Membrane platform's ability to dynamically create and run integration actions based on descriptions, which is a core feature of the service.
[PROMPT_INJECTION]: Indirect prompt injection surface identified:
Ingestion points: Data returned from Slash GraphQL via membrane action run commands.
Boundary markers: Absent in the prompt templates.
Capability inventory: Access to execute shell commands via the CLI and run dynamically created actions.
Sanitization: No explicit sanitization or filtering of external API responses is mentioned.

slash-graphql