close-automation
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting and processing external data during CRM automation tasks.
- Ingestion points: User-provided text fields for tool calls such as
CLOSE_CREATE_LEAD(name/contacts),CLOSE_CREATE_SMS(message text), andCLOSE_CREATE_TASK(description) inSKILL.mdworkflows. - Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions to separate potentially adversarial content in user inputs from the agent's system instructions.
- Capability inventory: The skill utilizes network-enabled tools to interact with the Close CRM API, including capabilities to send SMS messages, modify lead data, and delete call records.
- Sanitization: No logic is provided to sanitize, validate, or escape user-supplied strings before they are processed by the Rube MCP tools.
Audit Metadata