add-task

A benchmark task seed requires specific fields: discipline, task_id, task_name, description, inputs, outputs, standards, and complexity. The skill captures these through an expert interview and validates them against the seed-schema before file creation.

Yes, you can assess task feasibility before seed creation. The skill evaluates your captured task information against five feasibility criteria to determine if the task can be automated via templates, ensuring only viable seeds are written.

To structure inputs and outputs for seed generation, provide your task details to the skill. It extracts the required fields, validates them against the seed-schema, and writes a source_task.json file to the seeds directory for the generation workflow.

The best way to capture expert interview data for benchmark planning is using an interview-based skill. It extracts task fields like description, complexity, and standards, then evaluates feasibility and writes a structured source_task.json file.

You need a source_task.json file when you want to feed an expert's benchmark task idea into a generation pipeline. It serves as the structured seed payload containing all validated task fields and feasibility assessments required for automation.