skill-creator-workflow

Skill creation workflow applies a five-phase process to turn underspecified capability requests into rigorous, production-ready units. It uses scoping, facet decomposition, adaptive parallel research, synthesis, and final escalation with auditable artifacts.

The best way to orchestrate parallel research is defining focused research briefs, choosing sub-agent depth based on facet complexity, preserving findings on disk, and tolerating partial failures. This approach ensures robust research coordination within the skill creation workflow.

Quality governance checks for duplication, stack alignment, actionable instructions, reference density, and progressive disclosure before approval. These checks ensure correct agent attribution and validate the skill against existing stack components.

You can use this workflow for technical, workflow, methodological, integration, domain-expertise, and quality-gate skills. It requires scoped research, decomposition, synthesis, and orchestration approval for any of these skill types.

A Phase 5 escalation package is the final structured output of the skill creation workflow, prepared for orchestrator decision-making. It follows synthesis and drafting, presenting the research-backed skill for final quality governance and approval.