jinko-model
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill is a legitimate tool for working with the Jinkō modeling platform and follows security best practices.\n- [EXTERNAL_DOWNLOADS]: The skill instructions and scripts refer to the installation of the
jinko-sdkPython package and thejinko-skillsrepository from thenovainsilicoGitHub organization. These are vendor-provided resources necessary for the skill's functionality.\n- [INDIRECT_PROMPT_INJECTION]:\n - Ingestion points: The skill processes model definitions, ODE formulas, and tagging information provided by the user or read from existing models (e.g., in
SKILL.mdandscripts/edit_existing_model_batch.py).\n - Boundary markers: The instructions do not specify natural language delimiters for formulas, but the SDK enforces structural integrity through component-specific methods.\n
- Capability inventory: The skill uses
JinkoClientto create and modify project items, andmodel.simple_solve()to execute model simulations.\n - Sanitization: The skill mandates checking
model.diagnostics.errors(), setting strict unit-checking modes (UnitCheckAndConvertAllSpeciesToExtentUnits), and performing verification solves to ensure model sanity.\n- [COMMAND_EXECUTION]: The skill provides Python scripts (scripts/create_minimal_model.py,scripts/edit_existing_model_batch.py) and references CLI tools (jinko.cli.validate_model_readiness) to perform modeling tasks. These scripts include dry-run logic and use standard argument parsing to prevent unintended system or model mutations.
Audit Metadata