The Agent Skills Directory

[COMMAND_EXECUTION] (LOW): The skill contains instructions to execute various shell commands including pytest for contract testing, streamlit run for launching the UI, and recursive find commands to clean up Python cache files. These represent a standard developer execution surface.
[EXTERNAL_DOWNLOADS] (LOW): The guide references pip and uv for dependency management. While these tools can download external code, the instructions specifically target local installation (pip install -e .) and execution within the project environment.
[DYNAMIC_EXECUTION] (LOW): The troubleshooting guide mentions importlib.reload as a method for clearing module caches. While this is a form of dynamic loading, it is explicitly described as "not recommended" and is a standard Python development technique, thus the severity is minimized.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill uses "token efficiency" as a justification to steer the agent away from reading the original source code, encouraging reliance on the provided reference files instead. This creates a surface where the agent might adopt the skill's claims over the ground truth of the codebase.
Ingestion points: Reference markdown files provided in the skill.
Boundary markers: Absent; the agent is not instructed to verify the skill's reference files against the source.
Capability inventory: Subprocess execution (tests, app), file system cleanup (rm).
Sanitization: None; the agent is conditioned to trust the skill's mapping table to save tokens.

integrate-playground