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 and report_compiler_agent utilize the pdftotext command-line utility to extract the first lines and publication dates from PDF documents. This involves running shell commands on external files provided during the research process.
  • [EXTERNAL_DOWNLOADS]: The system connects to several well-known and established academic infrastructure services to verify references, including Semantic Scholar (api.semanticscholar.org), Crossref (api.crossref.org), and OpenAlex (api.openalex.org). These connections are used solely for bibliographic metadata retrieval and deduplication, which is essential to the skill's research function.
  • [SAFE]: The skill implements a highly structured multi-agent architecture with mandatory ethics and rigor checkpoints. It uses advanced citation markers (<!--ref:slug--> and <!--anchor:...-->) and applies percent-encoding to prevent data corruption or instruction leakage during the synthesis of external content. The internal logic is heavily focused on academic integrity and reproducibility.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 12:18 PM
Security Audit — agent-trust-hub — deep-research