deepwork_jobs

To define multi-step AI workflows, use an interactive process that gathers requirements to create a structured job.yml specification. This specification serves as the foundational blueprint for orchestrating automated AI tasks.

The best way to validate a job.yml specification is to review it against quality criteria using a sub-agent. This ensures unbiased validation and confirms the workflow meets structural and operational standards before implementation.

You implement step instructions by generating instruction files and syncing slash commands directly from the validated job.yml specification. This translates the workflow definition into executable actions for orchestration.

Automated learning improves multi-step AI workflow instructions by analyzing conversation history after a job runs. It captures learnings to refine future runs and updates the AGENTS.md pipeline for better orchestration.

No external dependencies are required to orchestrate and refine multi-step AI workflows. The system operates independently using the job.yml definition and AGENTS.md to manage structured automated learning pipelines.

You should use a structured job.yml definition for workflow automation when you need repeatable, auditable results. It provides a structured toolkit to coordinate multi-step AI tasks, ensuring consistent execution over manual prompting.