subagent-protocol

Parallel subagent dispatch across independent code domains requires defining explicit domain boundaries and detailed task specifications to ensure context isolation. The coordination uses a formal integration plan with stage gates to verify independence and prevent merge conflicts.

A two-stage review workflow enforces Stage 1 spec conformance and Stage 2 quality review gates before subagent integration. This mechanism ensures deterministic multi-agent work patterns across bounded domains while preserving verifiability.

Context isolation discipline is needed when tasks touch separate code domains with no shared mutable state. It ensures minimal sufficient context to prevent leaking unrelated state and avoid cross-domain dependencies during parallel dispatch.

Generating an integration plan and matrix requires explicit domain boundaries and feasibility checks for each task specification. The matrix verifies domain independence and prevents merge conflicts during post-dispatch verification.

Subagent coordination does not work for tasks with shared mutable state. This protocol applies strictly to tasks touching separate code domains to ensure isolation, requiring no shared mutable state for correct integration.

Limitations of two-stage review gates include the strict requirement for detailed task specifications and explicit domain boundaries. The protocol cannot integrate subagents without formal stage gates, success criteria, and post-dispatch verification.