novelty-check

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security issues detected. The skill uses tools that are appropriate for its stated purpose of academic novelty verification.
  • [DATA_EXFILTRATION]: The skill transmits the research idea provided in the arguments to external search engines via the WebSearch tool. This is a functional requirement for checking literature and is not considered malicious exfiltration in this context.
  • [PROMPT_INJECTION]: The skill fetches and analyzes external research papers via the WebFetch tool, which introduces a potential surface for indirect prompt injection.
  • Ingestion points: External research papers and abstracts are retrieved via WebFetch in Phase B (step 3) and Phase C (step 1).
  • Boundary markers: None identified; the agent is instructed to read the full text of retrieved documents.
  • Capability inventory: The skill is limited to search and file-reading tools (WebSearch, WebFetch, Grep, Read, Glob) and does not have access to shell execution or persistent storage modification tools.
  • Sanitization: No specific sanitization or filtering of fetched content is performed before analysis. Given the skill's specific use-case and limited tool set, this surface risk is acceptable.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 03:11 AM
Security Audit — agent-trust-hub — novelty-check