inworld-ai
Pass
Audited by Gen Agent Trust Hub on Apr 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the official
@membranehq/clipackage from the public npm registry, which is a required tool for interacting with the Membrane ecosystem. - [COMMAND_EXECUTION]: The skill executes the
membranecommand-line interface to manage service connections, perform authentication, and run actions against the Inworld AI platform. - [PROMPT_INJECTION]: The skill ingests action definitions and schemas via the
membrane action listcommand, creating a surface for indirect prompt injection if external metadata contains malicious instructions. - Ingestion points: Output from the
membrane action listcommand which provides action IDs and input schemas. - Boundary markers: None identified in the instruction set.
- Capability inventory: Execution of arbitrary actions and raw HTTP requests via the
membrane action runandmembrane requestcommands. - Sanitization: The skill relies on the platform to provide valid schemas and does not perform independent validation or escaping of the retrieved metadata.
Audit Metadata