return-refund-analyzer
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates in a read-only capacity, focusing on data aggregation and business intelligence. No evidence of malicious code, obfuscation, persistence, or privilege escalation was found.\n- [EXTERNAL_DOWNLOADS]: Data retrieval is restricted to the vendor's official Model Context Protocol (MCP) endpoint.\n
- Evidence: The skill references
https://mcp.datadoe.com/mcp/v1, which is the legitimate platform domain for DataDoe (a Deltologic resource).\n- [COMMAND_EXECUTION]: Analysis is performed using structured tool calls (e.g.,exports_create,sellers_and_vendors_list). There is no reliance on shell commands or dynamic execution of untrusted input.\n- [PROMPT_INJECTION]: The skill instructions define specific logic for bucketing return reasons and computing rates, which acts as a structured framework that mitigates the risk of obedience to external data instructions.
Audit Metadata