Multi-agent parallel implementation that turns an architecture document into working code. Spawns executor subagents for independent components and gates with reviewer. Use for the BUILD phase.
<Use_When>
/build command is invoked.omc/artifacts/03-architect/architecture.md exists and is non-empty<Do_Not_Use_When>
.omc/artifacts/03-architect/architecture.md — extract tech stack, file structure, data model, API contracts, coverage matrix.omc/artifacts/02-plan/prd.md — extract task priorities (P0/P1/P2)knowledge/claude-code-patterns/agent-design.md, , knowledge/claude-code-patterns/context-management.mdknowledge/claude-code-patterns/team-coordination.md.omc/state/pipeline-state.json if it existsFrom the coverage matrix and dependency graph:
Identify independent components that can be built in parallel.
For each component group:
If components are independent (no shared state or sequential dependency):
If components are dependent:
Track progress per component: not-started → in-progress → complete → reviewed.
After all components are implemented:
Invoke the reviewer agent to audit the implementation:
Assemble with these exact H2 sections:
## Implementation Summary (what was built, architecture changes, file count)## Component Status (table: component, status, justification)## File Manifest (all files created/modified with descriptions)## Build Verification (compile/run evidence, dependency status)## Review Status (reviewer assessment)## Deferred Items (what wasn't built and why)## Architecture Coverage (component → implementation mapping)Write to .omc/artifacts/04-build/build-report.md. Update pipeline state.
<Tool_Usage>
<Escalation_And_Stop_Conditions>
/architect first<Final_Checklist>
.omc/artifacts/04-build/build-report.md