adaptyv
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to install established scientific Python packages from official registries, such as
requests,pandas,biopython, andfair-esm. It also references interactions with legitimate bioinformatics web services, including the Adaptyv API and the NetSolP service hosted by the Technical University of Denmark (DTU). - [COMMAND_EXECUTION]: The skill provides examples of local file management for scientific workflows, such as creating directories and writing experimental data to JSON and CSV formats using Python's standard libraries.
- [DATA_EXFILTRATION]: The skill facilitates the transmission of protein sequence data to the Adaptyv Bio API for the purpose of experimental testing. This behavior is the primary function of the skill and is protected by standard Bearer token authentication.
- [PROMPT_INJECTION]: The skill ingests data from external API responses, representing a surface for indirect prompt injection. No malicious content was found in the static definitions.
- Ingestion points: Data retrieved from the Adaptyv API via endpoints for experiment results and target catalog searches (e.g., in
reference/examples.md). - Boundary markers: No explicit delimiters for untrusted data were identified in the provided implementation examples.
- Capability inventory: The skill can perform local file writes and display content to the user.
- Sanitization: No specific sanitization of API response content was observed in the examples.
- [SAFE]: No security issues or malicious behaviors were detected. The skill is consistent with its stated scientific purpose.
Audit Metadata