The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the @membranehq/cli package from the official NPM registry. This is the standard command-line interface provided by the vendor for interacting with their service.
[COMMAND_EXECUTION]: The skill uses local shell commands (membrane login, membrane connect, membrane action run) to manage the integration. These commands are part of the official vendor toolkit and are used as intended for service orchestration.
[CREDENTIALS_UNSAFE]: The skill demonstrates a strong security posture by explicitly stating that users should never be asked for API keys or tokens. Instead, it uses a connection-based model where credentials are managed server-side by the Membrane platform.
[INDIRECT_PROMPT_INJECTION]: The skill involves processing data from external actions (membrane action run). While this represents a data ingestion surface, the skill is a standard integration tool and does not exhibit high-risk autonomous behavior or lack of oversight that would escalate the concern beyond a baseline risk factor.

sailthru