The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages Jinkō trial visualizations through a typed SDK. It encourages the use of python-dotenv for secure environment variable management and does not exhibit patterns of unauthorized data exfiltration or access to sensitive local paths.
[COMMAND_EXECUTION]: The bundled Python script scripts/trial_viz.py provides a controlled CLI interface for SDK operations. It implements safety features such as input validation via regular expressions for labels and a dry-run pattern that requires an explicit --apply flag for any operation that modifies project state.
[INDIRECT_PROMPT_INJECTION]: The skill processes trial visualization JSON data, which is an ingestion surface for indirect prompt injection. However, the risk is mitigated by the structured nature of the data and the use of typed SDK models to handle the visualization content.
[DYNAMIC_EXECUTION]: The utility script uses dynamic loading to verify the presence of the jinko SDK and python-dotenv library. This is a standard and safe implementation for checking optional dependencies and does not involve the execution of untrusted code.

jinko-trial-viz