gza-task-debug

To diagnose gza task failures, analyze task logs, detect repetitive agent loops, and compare current metrics against recent baselines. This process surfaces root causes and patterns from logs, diffs, and cost metrics to deliver a structured diagnosis.

Detect agent loops and thrashing in task execution by analyzing task logs for repetitive patterns. This diagnostic approach identifies cyclical behaviors that cause tasks to fail or hit max_turns limits, surfacing the root cause of the execution failure.

The best way to quantify metric drift is to compare current task metrics against recent baselines. By fetching baseline statistics and evaluating differences in logs and cost metrics, you can accurately measure performance drift and identify anomalies.

Yes, you can use Python-based tooling to fetch task details, baseline statistics, and git branch information. This approach gathers the necessary execution context to produce a structured diagnosis with recommended remediation steps.

A gza task hits max_turns without completing due to repetitive loops or thrashing in task execution. Analyzing logs and comparing task metrics against baselines surfaces these execution patterns, identifying the root cause of the incomplete run.

To generate a diagnostic report for a failed task, you need to provide the task ID of the failed or max_turns task. This input triggers log analysis, loop detection, and baseline comparison to output actionable remediation recommendations.