tpmi-phewas-skill
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- File System Write Capability: The script
scripts/tpmi_phewas.pyallows saving data to a file path provided via theraw_output_pathparameter. - Evidence: The script uses
Path(raw_output_path).expanduser()andpath.write_text()to save API response data to a location specified in the input JSON. - Context: This feature is intended for exporting large data payloads that are too big for direct chat display. Because the path is expanded and written to without strict path validation, it is a security consideration regarding file system integrity. The content written is limited to JSON data received from the remote genomic API.
- External Service Interaction: The skill performs network requests to external genomic data providers to resolve variant coordinates and fetch association summaries.
- Evidence: The code makes HTTP GET requests to
pheweb.ibms.sinica.edu.twandrest.ensembl.orgusing therequestslibrary. - Context: These are recognized scientific resources in the bioinformatics community. Data retrieval is a core function of this skill.
- Indirect Prompt Injection Surface: The skill processes and summarizes data retrieved from external APIs, creating a potential surface for indirect instructions if a remote source were compromised.
- Ingestion points: Data is ingested from the TPMI and Ensembl REST APIs in
scripts/tpmi_phewas.pyandscripts/variant_resolution.py. - Boundary markers: The skill instructs the agent to provide concise summaries and use specific JSON structures.
- Capability inventory: The skill has the ability to make network requests and write files to the system.
- Sanitization: The script ensures the retrieved data is valid JSON before processing, but does not include explicit filtering for natural language instructions within the genomic data fields.
Audit Metadata