qinghu-1688-sourcing
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill interacts with an external API at
https://www.iqinghu.com/api/desktop/qh/mcpto fetch product data. It utilizes authentication credentials managed through environment variables (QINGHU_TOKEN,QHKIT_TOKEN) or direct user input. - [COMMAND_EXECUTION]: The instructions guide the agent to use standard tools like
curl,fetch, orrequeststo communicate with the vendor's HTTP endpoints, which is a routine operation for this type of integration. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes product titles, descriptions, and supplier information from the 1688 marketplace, which introduces a potential surface for indirect prompt injection from third-party content.
- Ingestion points: Data is parsed from the
content[0].textfield of the vendor API response as described in theSKILL.mdfile. - Boundary markers: The instructions do not specify any delimiters or safety prompts to ignore instructions that might be embedded within the retrieved product descriptions.
- Capability inventory: The skill has the capability to make further network requests to the vendor API and generate formatted reports based on the ingested data.
- Sanitization: No explicit validation or sanitization mechanisms are defined for the text content received from the 1688 platform.
Audit Metadata