The Agent Skills Directory

Human-in-the-Loop Controls: The skill implements multiple 'CRITICAL INTERACTIVE CHECKPOINT' steps. These require the agent to pause and obtain explicit user confirmation via a structured tool before proceeding with configuration, code generation, or script execution. This is an excellent safety practice that prevents autonomous execution of potentially unintended commands.- Controlled Code Generation: The skill uses a predefined Python script template to generate its execution logic. This limits the scope of generated code to the intended functionality of loading the Meridian model and serializing the output.- Use of Trusted External Resources: The skill provides a link to an official Google Colab notebook (colab.research.google.com) for data visualization. This link points to a well-known service and an official repository associated with the Meridian project.- Data Handling Practices: The skill correctly identifies binary files (.binpb) and instructs the agent to use proper loading methods (meridian_serde) rather than attempting to read them as text, which avoids potential processing errors.

meridian-scenario-planner