reference-corpus-analyzer

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious code, obfuscation, or unauthorized access patterns were identified. The skill operations are entirely consistent with academic literature analysis.\n- [PROMPT_INJECTION]: The skill processes external data from academic papers, which constitutes an indirect prompt injection surface.\n
  • Ingestion points: Raw papers fetched via WebSearch/WebFetch, source cards, and the project's source index.\n
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing external paper content.\n
  • Capability inventory: The skill uses Bash, Write, Edit, and WebFetch to manage analysis and synthesis tasks.\n
  • Sanitization: There are no explicit content sanitization or validation steps mentioned for the literature data being processed.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 07:22 PM
Security Audit — agent-trust-hub — reference-corpus-analyzer