Follow a tool call from proposal to execution. Fix missing hooks, inspect a trace, and test a policy before it can block a run.
Add a policy decision before a Python tool executes. Trace proposals separately from execution, test Audit mode, and handle unavailable Guard decisions.
Choose the right hook boundary for Claude Code tracing and blocking. Understand background execution, inline checks, and the evidence each produces.
Diagnose missing Codex traces by checking hook trust, project scope, the Python path, and the first real agent event.
Run a LiteLLM function-calling example with a local order lookup. Follow the model proposal, Python execution, and final answer in a Staso trace.