Deploy Persona-Specialized Agent Teams

You know how to dispatch agent swarms. Agent Teams adds persona specialization, inter-agent messaging, and proactive conflict detection.

Enable Agent Teams

Add the team block to your project's .claude/settings.json. This switches SPOQ from generic agent swarms to persona-specialized workers with inter-agent messaging and conflict detection.

.claude/settings.json

{
  "permissions": {
    "allow": [
      "Bash(npm test:*)",
      "Bash(npx next build)"
    ]
  },
  "team": {
    "enable": true,
    "personas": "default"
  }
}

"enable": true — activates persona assignment, inter-agent messaging, and conflict detection
"personas": "default" — uses the built-in 6 personas (Frontend, Backend, Infrastructure, Testing, Documentation, Fullstack)

Dispatch Your First Team

Same epic, new command. Here's how /team-execution differs from /agent-execution:

Enable Teams

Add "team": { "enable": true } to your project's .claude/settings.json. This tells SPOQ to use persona-specialized workers instead of generic agents.

cat .claude/settings.json

{
  "permissions": { ... },
  "team": {
    "enable": true,
    "personas": "default"
  }
}

Dispatch

Run /team-execution instead of /agent-execution. SPOQ computes waves, assigns personas to each task based on required skills and file patterns, and detects shared-file conflicts before spawning workers.

/team-execution @spoq/epics/config-validator

Team Execution Started
═══════════════════════════════
Personas loaded: Frontend, Backend, Infrastructure, Testing, Documentation, Fullstack
Conflict scan: Header.tsx, Footer.tsx flagged (shared across T01, T03)

Wave 1/3 — Dispatching 2 persona workers
  [T01] Frontend:  Define component types ... DONE (2m 31s)
  [T02] Backend:   Create validation utils .. DONE (3m 12s)

Wave 2/3 — Dispatching 2 persona workers
  [T03] Fullstack: Build CLI parser ......... DONE (4m 05s)
  [T04] Testing:   Add error formatters ..... DONE (2m 58s)

Wave 3/3 — Dispatching 2 persona workers
  [T05] Testing:   Integration tests ........ DONE (3m 41s)
  [T06] Docs:      Documentation ............ DONE (1m 28s)

Orchestrator: merging shared files (Header.tsx, Footer.tsx)
All 6 tasks completed in 3 waves
Total wall-clock time: 10m 22s
Speedup: 5.3x

Rework

When validation flags an issue, SPOQ sends feedback directly to the existing worker via inter-agent messaging. The worker retains full execution context — no respawn, no lost state.

# Automatic — no manual step needed

Code Validation: T03 scored 82/100 (below 95 threshold)
─────────────────────────────
Sending rework feedback to Fullstack worker...
  Context preserved: 4m 05s of prior execution
  Feedback: "Missing null check in CLI parser for empty args"

Fullstack worker: rework complete (1m 12s)
Re-validation: T03 scored 97/100 — PASS

Graceful shutdown: 6/6 workers confirmed
Team dissolved.

Agent Teams vs Basic Execution

How persona-specialized teams improve on the standard agent execution model.

Aspect	Basic Execution	Agent Teams
Agent Model	Generic Opus workers	Persona-specialized Opus workers
Rework Strategy	Spawn new agent (loses context)	Message existing agent (preserves context)
Shared Files	Implicit conflicts, manual merge	Proactive detection, orchestrator merge
Communication	TaskOutput polling	SendMessage with auto-delivery
Persona Support	None	6 domain-specific personas
Observability	TaskOutput polling	TaskList + automatic message delivery
Shutdown	Agents exit when done	Graceful shutdown via shutdown_request

Persona-Specialized Workers

Each agent receives a domain-specific prompt with coding standards, file pattern awareness, and trigger skills matched to the task requirements.

Frontend

React, Next.js, Tailwind CSS, accessibility, and responsive design

Skills: react, nextjs, tailwind, css, a11y
Files: src/components/ *.tsx *.css

Backend

Java, Spring, Node.js, Express, PostgreSQL, and API design

Skills: java, spring, node, express, postgresql, api
Files: src/service/ src/api/ *.java *.py

Infrastructure

Terraform, Docker, AWS, Kubernetes, and CI/CD pipelines

Skills: terraform, docker, aws, kubernetes, cicd
Files: infrastructure/ *.tf Dockerfile

Testing

Jest, Playwright, Pytest, Vitest, and end-to-end testing

Skills: jest, playwright, pytest, vitest, e2e
Files: *.test.* tests/ e2e/

Documentation

LaTeX, Markdown, technical writing, and SEO optimization

Skills: latex, markdown, technical-writing, seo
Files: documents/ *.tex *.md

Fullstack

Catch-all for tasks spanning multiple domains or requiring broad expertise

Skills: fullstack, cross-domain, integration
Files: *

Proven at Scale

Real-world benchmarks from 9 project deployments using SPOQ.

1.3–5.3x

Speedup Range

Faster than sequential single-agent execution

147

Tasks Validated

Across 9 real-world project deployments

100%

Completion Rate

All tasks passed dual validation gates

0.90–0.95

Agent Confidence

Self-reported confidence scores

6 tasks/hr

Peak Velocity

Sustained throughput during parallel waves