jinko-task-export-to-matlab

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests simulation workbooks (.xlsx) and configuration metadata (.yaml) extracted from model bundles, providing a surface for data-embedded instructions.
  • Ingestion points: The helper script scripts/jinko_matlab_parity.py extracts files from model bundles and reads simulation settings.
  • Boundary markers: The instructions lack explicit delimiters or warnings to ignore natural language prompts found within the data.
  • Capability inventory: The agent is capable of writing to the local filesystem, performing network operations via the Jinko SDK, and executing MATLAB code.
  • Sanitization: The validate_settings function provides robust validation of numeric parameters, verifying they are finite and within expected bounds.
  • [COMMAND_EXECUTION]: The workflow involves generating and executing MATLAB scripts (sbiosimulate) based on parameters extracted from external files. This is managed through a deterministic helper script to reduce the surface area for errors.
  • [EXTERNAL_DOWNLOADS]: The skill fetches model bundles and settings from the Jinko platform (jinko.ai), which is the vendor's official infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 09:59 AM
Security Audit — agent-trust-hub — jinko-task-export-to-matlab