cnki-paper-detail
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs metadata extraction from academic papers on the CNKI platform. It uses standard browser tools for navigation and DOM parsing.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted content from external web pages into the agent's context. This is a common characteristic of web-scraping skills.
- Ingestion points: The skill reads page content using
browser action=evaluateandbrowser action=snapshot(SKILL.md, Step 3 and 5). - Boundary markers: While the output is structured into distinct markdown sections (Step 4), it lacks explicit isolation delimiters or instructions for the agent to ignore potentially malicious text embedded within the paper's title or abstract.
- Capability inventory: The agent has access to browser automation capabilities, including navigation, script execution, and snapshotting.
- Sanitization: The scraping script uses
.innerTextto retrieve text, which effectively prevents HTML/Script injection but does not mitigate instructions contained in the text itself. - [SAFE]: The JavaScript used for scraping is hardcoded within the skill and does not download or execute remote code. All data extraction is performed on the client-side within the browser session.
Audit Metadata