trace

To trace agent execution flow, invoke trace_timeline to display a chronological sequence of how hooks, keywords, skills, agents, and tools interacted. This presents events in the exact order they occurred for diagnosis and optimization workflows.

You can view aggregated statistics for agent-driven sessions by running the trace_summary tool. It provides aggregated counts and a quick overview of session activity, helping you analyze interactions and performance over time.

To debug workflow divergences or slowdowns, run trace_timeline to review the latest session event sequence. This chronological trace highlights exactly where hooks, skills, or tools interacted poorly, enabling targeted debugging.

The best way to get a quick overview of recent session activity is to run trace_timeline with no arguments to display the latest session, followed by trace_summary to view aggregated counts and statistics.

Yes, filters are available to tailor agent flow analysis. The trace functionality supports filter and last parameters, allowing you to customize the timeline and summary outputs for targeted debugging and workflow optimization.

You need session tracing when diagnosing complex agent interactions in debugging and optimization workflows. It helps you understand how components interacted chronologically and identify where the flow diverged or slowed during a session.