drzero-swarm
Distribute work across multiple domain specialist agents in parallel for complex multi-domain tasks
DrZero Swarm Mode
Distribute work across multiple domain specialists simultaneously. The orchestrator coordinates parallel execution and synthesizes results.
Installation Prerequisite
Install the DrZero Codex bundle before using this skill:
make codex-install-drzero SCOPE=repo
Workflow
1. Task Decomposition
- Dispatch agent: drzero-orchestrator
- Break complex task into domain-specific subtasks
- Identify dependencies between subtasks
- Plan execution order (parallel where possible)
2. Parallel Dispatch
- Launch independent subtasks in parallel (up to 6 concurrent)
- Codex
agents.max_threads = 6controls parallelism - Each subtask dispatches to its domain specialist
3. Dependency Resolution
- Wait for prerequisite subtasks to complete
- Pass outputs from completed tasks to dependent tasks
- Handle failures with retry or escalation
4. Result Synthesis
- Orchestrator collects all specialist outputs
- Merge changes (resolve conflicts if any)
- Generate unified summary
5. Review
- Dispatch agent: drzero-security on combined output
- Verify cross-domain consistency
- Run integration checks
Orchestration Patterns
Sequential Chain
Task A -> Task B -> Task C (each depends on prior)
Parallel Fan-Out
Task A, Task B, Task C all run simultaneously, then merge
Bounded Batch
Process N items with max M concurrent (e.g., 20 files, 6 at a time)
Coordination Detail
Domain Routing
The orchestrator maps each subtask to one of 16 canonical domain specialist agents (e.g.,
drzero-architecture, drzero-security, drzero-performance, drzero-testing,
drzero-documentation, drzero-database, drzero-frontend, drzero-backend, etc.).
Routing uses keyword matching on the subtask description and file-path heuristics
(e.g., files under tests/ route to the testing specialist).
Parallel Execution Strategy
- The orchestrator builds a dependency DAG from the decomposed subtasks.
- Independent subtasks (no shared file writes) are dispatched immediately, up to the
parallel_threadslimit (default: 6, configurable viadrzero-config). - Dependent subtasks wait in a ready queue until prerequisites complete.
- If a specialist fails or times out, the orchestrator retries once before marking the subtask as failed and continuing with remaining work.
Conflict Resolution
When two specialists propose changes to the same file:
- The orchestrator detects the overlap during result collection.
- It applies changes sequentially in priority order (security > architecture > other).
- If the second patch cannot apply cleanly, the orchestrator re-dispatches the lower-priority specialist with the updated file content as context.
Result Aggregation
After all subtasks complete, the orchestrator:
- Merges file-level diffs into a unified changeset.
- Runs the drzero-security agent on the combined output for cross-domain review.
- Produces a summary listing each specialist's contribution, files changed, and any conflicts that were resolved.
Use Cases
- Multi-component refactoring (architecture + implementation + testing)
- Full-stack feature (frontend + backend + database + infrastructure)
- Cross-cutting concerns (security + compliance + documentation)
- Large-scale migrations (process files in batches)
Related Assets
Wall-E Agent Composition Helper
Compose multiple specialized agents into a safe Wall-E workflow with proper MCP tool assignments, guardrails, and human-in-loop gates.
Owner: epic-platform-sre
Wall-E Workflow Designer (Optum)
Assist with designing, reviewing, and optimizing multi-agent Wall-E workflows and MCP integrations following Optum enterprise patterns.
Owner: epic-platform-sre
Wall-E Orchestration Patterns (Optum)
Patterns and guardrails for composing safe multi-agent workflows in Wall-E (Wide Array Large Language Engine), Optum's enterprise AI orchestration platform.
Owner: epic-platform-sre
drzero-unity
Coordinate peer-to-peer DrZero parallel work without a single central implementation owner
Owner: epic-platform-sre
drzero
Run DrZero autonomous improvement after initializing shared drzero.yml while preserving the @drzero plugin mention
Owner: epic-platform-sre
drzero-analysis
Deep codebase analysis without making changes - architecture review, quality assessment, and improvement recommendations
Owner: epic-platform-sre

