deep-research

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The timeline_extraction_agent is designed to execute the 'pdftotext' command-line utility locally. It uses this tool to extract publication dates from the first page of user-provided or retrieved PDF files by running 'pdftotext -f 1 -l 1 '.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates metadata retrieval and source verification by making network requests to established academic services. These include Semantic Scholar (api.semanticscholar.org), Crossref (api.crossref.org), OpenAlex (api.openalex.org), and arXiv (export.arxiv.org).
  • [SAFE]: The skill documentation includes a 'canonical:instruction-data-boundary' protocol. This directive explicitly instructs the agents to treat all retrieved external content (web pages, PDFs, and third-party documents) as data rather than instructions. This architecture effectively shields the system from indirect prompt injection attempts where malicious commands might be hidden in fetched research materials.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 11:30 AM
Security Audit — agent-trust-hub — deep-research