kata-agent-specialists-design

Designing specialist sub-agents for workflow orchestration requires converting an orchestrator's specialist list and handoff states into bounded specialist markdown artifacts, ensuring safe delegation and clear scope.

To avoid scope overlap when building specialist agents, you must enforce explicit no overlap constraints across their responsibilities. This prevents duplicated scope and inconsistent tool usage by validating aggregate boundaries during the agent design phase.

You should delegate aggregate boundary design to a domain model authority whenever a specialist operates over a canonical entity in documentation. This ensures aggregate boundary correctness by validating the mapping evaluation before generating the specialist directly.

Generating markdown documentation for agent specialists involves producing specialists/{name}.md files in the correct docs path. This output references or updates docs/{context}/entities when domain model aggregates are created or adjusted during delegation.

The best way to map orchestrator handoff states to specialist agents is by reading the orchestrator's declared specialists and state transitions to align boundaries. This ensures each specialist either maps to a canonical docs entity or represents a distinct cross-cutting technical capability.