alibaba-product-research
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it ingests untrusted data from a public marketplace.
- Ingestion points: Product titles, supplier names, and attributes are extracted from Alibaba.com search results in
SKILL.md(Step 4) usingbrowser_evaluate. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the extracted data as untrusted or to ignore instructions embedded within the text.
- Capability inventory: The skill uses the Playwright MCP toolset (
browser_navigate,browser_click,browser_evaluate, etc.) to interact with the web. - Sanitization: The extraction script in
SKILL.mduses asafeTexthelper, but it only handles DOM accessibility and whitespace trimming. There is no sanitization or filtering of the actual content to prevent malicious instructions from being interpreted during the report generation phase (Step 6). - [SAFE]: The skill navigates to and fetches data from Alibaba.com, which is a well-known international commerce platform. These operations are essential to the skill's primary purpose and do not constitute a security risk.
Audit Metadata