pmc-official-download

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with official NCBI (National Center for Biotechnology Information) domains to retrieve research literature.
  • Evidence: The scripts scripts/download_pmc_bioc.py and scripts/download_selected_pmc_pdfs.py make network requests to ncbi.nlm.nih.gov subdomains (e.g., www.ncbi.nlm.nih.gov, eutils.ncbi.nlm.nih.gov, and ftp.ncbi.nlm.nih.gov).
  • Context: These are well-known, authoritative services for medical and biological research data. The use of these services is entirely consistent with the skill's primary purpose.
  • [COMMAND_EXECUTION]: The skill instructs the agent to run local Python scripts to automate the retrieval process.
  • Evidence: SKILL.md contains command-line instructions such as python scripts/download_pmc.py --pmcid PMC13011518.
  • Context: The included scripts use standard libraries (urllib, argparse, pathlib) and perform transparent operations. No suspicious subprocess spawning or privilege escalation attempts were found.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests and processes content from external academic articles.
  • Ingestion points: scripts/download_pmc_bioc.py and scripts/download_selected_pmc_pdfs.py download full-text XML, JSON, and PDF files to the local pmc-downloads/ directory.
  • Boundary markers: The skill does not implement specific boundary markers or "ignore instructions" wrappers for the content it downloads.
  • Capability inventory: The agent has the capability to write files to the disk and read network data via the provided scripts.
  • Sanitization: The scripts include a slugify function to sanitize file names derived from article titles, although identifiers like PMCIDs are used directly in file paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 04:47 PM