cnki-paper-detail
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts content from external web pages, which is an inherent surface for indirect prompt injection attacks if the web content contains malicious instructions designed to influence the AI agent. 1. Ingestion points: Data extracted from CNKI article pages via browser devtools tools. 2. Boundary markers: The instructions do not define clear delimiters or warning markers to isolate the extracted paper content from the agent's control logic. 3. Capability inventory: Uses mcp__chrome-devtools for navigation, DOM snapshotting, and script evaluation within the browser context. 4. Sanitization: Extracted text is trimmed and cleaned of specific UI markers (like 'Online First' labels) but does not undergo security-focused sanitization before being processed by the agent.
- [SAFE]: The skill's behavior matches its description. It uses specific DOM selectors and standard extraction techniques for its stated purpose of academic research assistance. No malicious persistence, exfiltration, or obfuscation was detected.
Audit Metadata