reference-search
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes external literature metadata from PubMed API responses.
- Ingestion points: XML and JSON data received from eutils.ncbi.nlm.nih.gov within scripts/pubmed_search.py.
- Boundary markers: Data is structured into JSON and CSV files using templates in the assets/ directory.
- Capability inventory: Network access via urllib.request and file system writing to the outputs/ directory.
- Sanitization: Outgoing queries are URL-encoded and results are parsed using standard library tools.
- [EXTERNAL_DOWNLOADS]: The script scripts/pubmed_search.py communicates with eutils.ncbi.nlm.nih.gov to fetch metadata. This is a well-known scientific service and is the primary data source for the skill.
- [COMMAND_EXECUTION]: Example instructions in SKILL.md include standard Python commands for script compilation and execution which are benign for this use case.
Audit Metadata