krayin
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include the installation of the
@membranehq/clipackage from the npm registry. This is the official command-line interface for the Membrane platform mentioned in the skill. - [COMMAND_EXECUTION]: The skill utilizes shell commands through the
membraneCLI to manage authentication, establish connections, and perform API operations. These commands are executed within the agent's environment to interact with the Krayin CRM. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process data from Krayin CRM records, which constitutes an untrusted data source.
- Ingestion points: Data retrieved from Krayin via
membrane action runandmembrane requestcommands (e.g., Leads, Deals, Organizations). - Boundary markers: There are no specific delimiters or instructions provided to the agent to treat the retrieved CRM data as untrusted or to ignore instructions embedded within it.
- Capability inventory: The skill possesses the capability to execute shell commands and make further network requests via the Membrane proxy based on agent logic.
- Sanitization: No explicit sanitization, validation, or escaping of the CRM data is described before the agent processes or interprets the content.
Audit Metadata