skills/az9713/cerebro/arxiv-analysis/Gen Agent Trust Hub

arxiv-analysis

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The skill's behavior is consistent with its stated purpose of academic research analysis.
  • [PROMPT_INJECTION]: Indirect prompt injection surface analyzed due to the ingestion of external web content.
  • Ingestion points: The skill fetches paper content from arxiv.org using the WebFetch tool (SKILL.md, Step 2).
  • Boundary markers: The instructions do not explicitly implement delimiter-based isolation for the fetched content.
  • Capability inventory: The agent is authorized to write analysis reports to reports/papers/ and update logs in logs/ (SKILL.md, Steps 7 and 8).
  • Sanitization: The skill performs basic sanitization on paper titles (lowercase and hyphenation) for safe file naming.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 09:59 AM
Security Audit — agent-trust-hub — arxiv-analysis