paper-summary

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data which introduces a surface for indirect prompt injection attacks.
  • Ingestion points: PDF files, URL links, and pasted text content as described in SKILL.md.
  • Boundary markers: The instructions do not define delimiters or specific markers to isolate the user-provided content from the agent's instructional context.
  • Capability inventory: The workflow utilizes tools such as extract_pdfs_full_content and extract_pdfs_key_info to ingest data.
  • Sanitization: There is no evidence of sanitization, filtering, or validation of the content extracted from the external sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 07:46 PM
Security Audit — agent-trust-hub — paper-summary