analyze-traj

name

analyze-traj

description

Analyze OSWorld-V2 agent trajectory logs and task results to produce actionable insights. Use this skill whenever the user wants to understand agent performance on OSWorld tasks — including analyzing trajectories, reviewing task results, finding error patterns, comparing code vs GUI strategies, identifying which tools/commands the agent used, or deciding which task types to scale up in the benchmark.

If only one task is issued, analyze it directly with instruction: analyze-single-traj.md.

If multiple tasks or a whole results directory are issued, use subagents to analyze them in parallel (one agent for each task). Do not analyze them sequentially by yourself. DO NOT tell it what to do. Just ask the subagent to analyze the task in target directory and use this skill (analyze-traj) to do the analysis. Pass any user instructions to every subagent.

After the per-task reports are ready:

• Do nothing but report to the user that the analysis is done and where to find the reports.
• Ask user if they want to synthesize a run-level summary, if yes use: analyze-full-run.md.