The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (design_experiments.py) to process research plans. The script is self-contained, uses only the Python standard library, and performs basic file I/O to read plan files and write experiment designs locally. There are no calls to external shells, network resources, or administrative commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes research ideas and plan files (research_plan.json) which are ingested into the agent context. While this constitutes a data ingestion surface, the transformation logic is limited to structured data generation (JSON/Markdown) for research planning and does not grant the skill access to sensitive capabilities or bypass agent safety constraints.

experiment-design