Goals · Actions · Memory · Environment — design the agent loop on paper, then implement it in Python with tool calling. Inspired by Dr. Jules White’s Vanderbilt teaching (Coursera AI Agents courses).
Worksheet the four pillars for your use case, then map them to Python modules.
Decide → act → remember, with iteration caps and a terminal finish tool.
Writable goals, JSON Schema tool contracts, and an action registry.
Message roles, truncation, sandboxed execution, and hidden secrets.
End-to-end Project Helper: venv, modules, run checklist, customize for your tools.