research-batch

Automate batch research data collection by launching independent background agents for each item listed in your outline.yaml. This coordinates parallel data gathering, reducing manual coordination and overall processing time.

Resume batch research progress through a built-in completion tracking mechanism that skips completed items. The workflow automatically detects existing results in the execution.output_dir and continues processing only the remaining pending items.

Yes, an outline.yaml file containing your research items list and execution configuration is required. The workflow auto-locates this file in your current project to manage batch execution and define the output directory.

Batch execution manages parallel agents by dividing your items according to configured batches and items_per_agent limits. It launches web-search-agents in the background to concurrently gather research data.

Data validation is handled by a validate_json.py script located with the skill files. This script verifies field coverage to ensure the per-item results gathered by the agents meet your specified requirements.

Limitations include dependency on a properly structured outline.yaml for execution configuration and the necessity of a local validate_json.py script. Processing also requires background execution capabilities for parallel agent management.