eva-skill
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [Indirect Prompt Injection Surface]: The skill processes data from external REST API responses which could potentially contain instructions aimed at the agent.
- Ingestion points: The script
scripts/rest_request.pyingests and parses JSON or text responses from the configured API. - Boundary markers: No explicit delimiters or instructions to ignore embedded content are wrapped around the processed data.
- Capability inventory: The script can perform network operations and write files to the local system.
- Sanitization: Data is truncated via the
_compactfunction before being returned to the agent. - [File System Write Capability]: The
_save_raw_outputfunction inscripts/rest_request.pywrites API responses to a path provided in the input payload. This permits the agent to write data to paths it has permissions for within its environment. - [Dynamic Network Target Specification]: The script allows the
base_urlandpathfor REST requests to be set via input parameters. While intended for EBI services, this flexibility allows communication with any reachable network endpoint. - [External Service Dependency]: The skill fetches data from the European Bioinformatics Institute (EBI) at
ebi.ac.uk, which is a well-known scientific service provider.
Audit Metadata