The Agent Skills Directory

[SAFE]: The skill utilizes the shinka_run and shinka_models CLI tools for its core functionality. These tools are used in a controlled manner, relying on user-specified paths and environment-provided configurations.
[SAFE]: The workflow incorporates user feedback into task-specific system prompts. The skill suggests using configuration files for complex inputs, which mitigates risks associated with direct shell interpolation.
[SAFE]: All external references target official SakanaAI resources and academic documentation. There is no evidence of unauthorized remote code fetching, execution, or data exfiltration.
[SAFE]: A strict human-in-the-loop policy is enforced, ensuring the agent stops for user approval before initiating execution batches or applying new settings.

shinka-run