The Agent Skills Directory

Structured Command Execution: The skill is designed to generate and execute temporary Python scripts to probe runtime environments. It utilizes subprocess.run to gather system metadata (such as git status and package versions) and execute the probe scripts. This is a core feature of the skill's diagnostic purpose.
Controlled Environment Access: There are explicit 'gates' and rules regarding environment variables (e.g., OPENAI_API_KEY). The skill instructions strictly forbid reading or printing these variables without explicit user approval and identify them by name beforehand.
External Service Interaction: The skill facilitates 'live probes' to external APIs (specifically OpenAI). It includes detailed guidance on managing these interactions, such as using 'warm-up' runs and 'repeat-N' patterns to account for caching and latency, while requiring explicit user intent for any mutating or costly operations.
Dependency Management: The skill utilizes uv run python to ensure probes are executed within the correct repository context and environment, reducing the risk of accidental library conflicts or side effects from the host system.
Data Handling and Cleanup: Instructions emphasize the use of temporary directories for artifacts and mandate their deletion after the final report is generated, unless otherwise requested by the user, maintaining a clean state.

runtime-behavior-probe