experiment-detail-comparator
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads PDF files and metadata from the Zotero API (api.zotero.org) and user-provided URLs using both the curl utility and the Python requests library.
- [REMOTE_CODE_EXECUTION]: The scripts scripts/download_full_pdf.py and scripts/extract_pdf_tables.py contain logic to detect missing Python dependencies (requests, PyPDF2, pdfminer, and pdfplumber) and automatically install them using pip at runtime.
- [COMMAND_EXECUTION]: The workflow heavily relies on subprocess.run to invoke external system tools, including zotcli/zotero-cli for library searches and the NCBI Entrez Direct suite (esearch, efetch) for literature retrieval.
- [INDIRECT_PROMPT_INJECTION]: Ingestion points: run_comparison.py and scripts/download_full_pdf.py ingest untrusted PDF documents from external sources. Boundary markers: The skill does not implement delimiters or warnings to ignore instructions embedded within the research papers. Capability inventory: The skill has the ability to execute shell commands, write files, and perform network operations. Sanitization: No sanitization is applied to the paper content before it is processed into markdown and analyzed by the agent.
Audit Metadata