bioresearcher-pubmed-weekly
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches PubMed update files in XML format from the official National Center for Biotechnology Information (NCBI) FTP server at
ftp://ftp.ncbi.nlm.nih.gov/pubmed/updatefiles/. This is a well-known and trusted repository for biomedical literature. - [COMMAND_EXECUTION]: The main workflow script
scripts/pubmed_weekly.pyusessubprocess.runto execute a bundled Python parsing script (scripts/parse_updatefiles.py). This implementation uses a secure list-based argument structure and invokes the current Python interpreter, which is a standard method for modular script execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes external XML data, creating an attack surface for indirect prompt injection.
- Ingestion points: External XML/XML.gz files are downloaded from the NCBI FTP server and parsed in
scripts/parse_updatefiles.py. - Boundary markers: The skill does not implement explicit delimiters or warnings to the agent regarding the untrusted nature of the downloaded content.
- Capability inventory: The skill can execute subprocesses (internal scripts) and write to the filesystem (
combined.xlsx,summary.json). - Sanitization: While the parser extracts specific XML fields and flattens markup, it does not perform explicit sanitization of the extracted strings to prevent potential downstream injection attacks, although the risk is low due to the trusted nature of the data source.
Audit Metadata