session-investigator

Diagnose unexpected session endings by reading session.json metadata and history_*.json files to identify stop reasons, pending tool calls, and timing anomalies that reveal the specific failure patterns.

Correlate sub-agent traces with main session history logs by examining history_*.json files to align assistant tool_calls with user tool_results and timing channels, diagnosing conversation flow misalignments and loops.

Identify tool loops in fast-agent history JSON files by using jq queries to inspect history_*.json structures, matching assistant tool_calls against user tool_results, and analyzing timing channels for repeated patterns.

Yes, you can use jq queries to analyze timing anomalies in session.json and history_*.json files, extracting stop reasons, pending tool calls, and timing channel data to detect conversation flow gaps.

Inspect history_*.json files to diagnose fast-agent failures when sessions end unexpectedly, when tool loops occur, or when correlating sub-agent traces with main sessions to analyze conversation flow and timing anomalies.

Fast-agent sessions loop tool calls without completing due to misalignments between assistant tool_calls and user tool_results; examining history_*.json timing channels and stop reasons diagnoses these failed loop patterns.