raet

Pass

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the @membranehq/cli package from the npm registry. This is a vendor-provided tool used to facilitate communication with the Membrane platform.
  • [COMMAND_EXECUTION]: The skill instructs the agent to run various shell commands using the membrane CLI (e.g., login, connection, action, request). These are used to manage authentication and interact with the Raet API.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data retrieved from the external Raet API, creating a potential surface for instructions embedded in remote data to influence agent behavior.
  • Ingestion points: Data enters the agent's context through the output of membrane action list, membrane action run, and membrane request commands.
  • Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided in the skill to wrap the data returned by these commands.
  • Capability inventory: The agent has the capability to execute further shell commands, perform network operations via the CLI, and manage connections based on the data it processes.
  • Sanitization: No explicit sanitization, validation, or filtering of the API response content is mentioned before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 18, 2026, 03:09 PM