judgeme
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the
@membranehq/clipackage globally usingnpm. This utility is required to facilitate communication between the agent and the Membrane platform. - [COMMAND_EXECUTION]: The skill makes extensive use of the
membraneCLI to perform operations such as logging in, creating connections, and running specific Judge.me actions (e.g.,list-reviews,update-shop). These commands are standard for the tool's integration purposes. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, specifically product reviews and store information from Judge.me, which may contain untrusted content.
- Ingestion points: Data retrieved via
membrane action run(e.g.,list-reviews) or raw API requests viamembrane requestinSKILL.md. - Boundary markers: No specific delimiters or warnings for the agent to ignore instructions within the retrieved data are provided.
- Capability inventory: The skill includes network operations via the CLI and the ability to perform write operations (e.g.,
create-public-reply,update-shop) based on processed data. - Sanitization: There is no explicit sanitization or filtering logic described for handling external review text before it is returned to the agent context.
Audit Metadata