grabfood
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the
@membranehq/clipackage from npm, which is an official tool provided by the vendor to facilitate platform interactions. - [COMMAND_EXECUTION]: The skill uses shell commands to interact with the Membrane CLI for logging in, managing connections, and executing data operations.
- [REMOTE_CODE_EXECUTION]: The
membrane action createfeature allows the platform to generate and run code based on natural language descriptions, representing a core functional feature of the Membrane environment. - [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection by processing external data from GrabFood actions. Ingestion points: Data returned from
membrane action runcommands. Boundary markers: None present. Capability inventory: Execution of CLI commands for data management. Sanitization: No explicit sanitization or filtering of action output is described.
Audit Metadata