Skills
Navigate
skills/cookjohn/wos-skills/wos-paper-detail/Gen Agent Trust Hub

wos-paper-detail

Pass

Audited by Gen Agent Trust Hub on Sep 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Web of Science website (paper titles, abstracts, authors) and presents it to the agent, which could allow maliciously crafted academic content to influence the agent's behavior.
  • Ingestion points: The skill uses a JavaScript block in SKILL.md that accesses the DOM via document.querySelector, document.body.innerText, and textContent to extract external data.
  • Boundary markers: The "Present Results" step in SKILL.md lacks explicit boundary markers or instructions for the agent to ignore instructions embedded within the extracted data.
  • Capability inventory: The skill uses navigate_page and evaluate_script to interact with external web content.
  • Sanitization: The JavaScript uses textContent (which prevents HTML parsing) and trim(), and truncates specific metadata fields to 300 characters, providing some level of mitigation against extremely long payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 30, 2026, 01:43 AM