literature-research

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves academic data from trusted and well-known repositories.
  • Evidence: scripts/literature_deep_review.py and scripts/pubmed_literature_search.py fetch research metadata and abstracts from NCBI E-utilities (eutils.ncbi.nlm.nih.gov) and ClinicalTrials.gov. Additionally, scripts/literature_deep_review.py downloads full-text XML from PubMed Central (PMC) and checks the Unpaywall API (api.unpaywall.org) for free article versions.
  • [PROMPT_INJECTION]: The skill ingests untrusted text content from external research databases, creating an attack surface for indirect prompt injection if the resulting data is analyzed by an LLM.
  • Ingestion points: Research titles, abstracts, and full texts are fetched from external APIs in scripts/pubmed_literature_search.py and scripts/literature_deep_review.py.
  • Boundary markers: Absent; article content is retrieved and stored without specific delimiters or instructions to ignore embedded commands.
  • Capability inventory: The scripts are restricted to network GET requests to academic APIs and local file writing within the .tmp/ directory. No capabilities for code execution or system-level changes are present.
  • Sanitization: Content is parsed using standard XML and JSON libraries, but narrative text is not filtered for potential malicious instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 02:19 PM