This skill should be used when the user says "deploy a team", "spin up agents to work on this", "use all our agents", "coordinate specialists", or wants to break a large task into parallel sub-tasks handled by multiple domain experts simultaneously. Orchestrates Claude Code's experimental agent team system using the full bopen-tools specialist roster.
From bopen-toolsnpx claudepluginhub b-open-io/claude-plugins --plugin bopen-toolsThis skill uses the workspace's default tool permissions.
references/agent-roster.mdreferences/permissions-and-isolation.mdreferences/spawn-prompt-guide.mdSearches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides AI-assisted editing of real video footage: transcribe/plan cuts with Claude, execute via FFmpeg bash scripts, augment with Remotion/ElevenLabs/fal.ai, polish in Descript/CapCut.
Deploy a coordinated team of specialized agents from the bopen-tools kit using Claude Code's agent team system. Agents work in parallel on independent tasks and communicate through a shared task list and message bus.
Agent teams require this env var to be set:
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
Add to ~/.claude/settings.json:
{
"env": {
"CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
}
}
Without this, TeamCreate will fail.
mode on Every Agent SpawnWARNING: The default permission mode will block teammates waiting for permission prompts that never arrive, stalling the whole team.
Always set mode: "bypassPermissions" when spawning teammates:
Agent(
subagent_type: "bopen-tools:designer",
team_name: "feature-billing",
name: "designer",
mode: "bypassPermissions", # ← REQUIRED
prompt: "..."
)
See references/permissions-and-isolation.md for all mode options and worktree isolation.
| Agent | subagent_type | Best for |
|---|---|---|
| researcher | bopen-tools:researcher | Libraries, APIs, docs, competitive analysis |
| nextjs | bopen-tools:nextjs | Next.js, React, Vercel, RSC, app router |
| designer | bopen-tools:designer | UI, Tailwind, shadcn/ui, accessibility |
| agent-builder | bopen-tools:agent-builder | AI agent systems, LLM integration, Vercel AI SDK |
| database | bopen-tools:database | Schema, queries, PostgreSQL, Redis, Convex |
| integration-expert | bopen-tools:integration-expert | REST APIs, webhooks, third-party services |
| code-auditor | bopen-tools:code-auditor | Security review, vulnerability scanning |
| tester | bopen-tools:tester | Unit, integration, e2e tests, CI |
| documentation-writer | bopen-tools:documentation-writer | READMEs, API docs, PRDs, guides |
| devops | bopen-tools:devops | Vercel+Railway+Bun, CI/CD, monitoring |
| optimizer | bopen-tools:optimizer | Bundle analysis, Lighthouse, Core Web Vitals |
| architecture-reviewer | bopen-tools:architecture-reviewer | System design, refactoring strategy, tech debt |
| mobile | bopen-tools:mobile | React Native, Swift, Kotlin, Flutter |
| payments | bopen-tools:payments | Stripe, billing, financial transactions |
| marketer | bopen-tools:marketer | CRO, SEO, copy, launch strategy |
| legal | product-skills:legal | Privacy, compliance, ToS |
| mcp | bopen-tools:mcp | MCP server setup, config, diagnostics |
Full roster with per-agent skills to mention in spawn prompts: references/agent-roster.md
Before calling any tools, identify:
TeamCreate(
team_name: "feature-billing",
description: "Implement Stripe billing with UI, API, tests, and docs"
)
Set dependencies with addBlockedBy where order matters:
TaskCreate(
subject: "Design billing UI components",
description: "Create PricingCard, BillingHistory, UpgradeModal using shadcn/ui.
Repo: ~/code/myapp. Tailwind v4. Output: src/components/billing/.",
activeForm: "Designing billing UI"
) → id: "1"
TaskCreate(
subject: "Implement Stripe integration",
description: "Set up webhooks, subscription creation, customer portal.
Repo: ~/code/myapp. API routes in app/api/billing/.",
activeForm: "Implementing Stripe integration"
) → id: "2"
TaskCreate(
subject: "Write billing test suite",
description: "Vitest tests for all billing API routes and webhook handler.
Repo: ~/code/myapp. Tests in __tests__/billing/.",
activeForm: "Writing billing tests"
) → id: "3"
TaskUpdate(taskId: "3", addBlockedBy: ["2"]) # tests wait for Stripe impl
Agent(
subagent_type: "bopen-tools:designer",
team_name: "feature-billing",
name: "designer",
mode: "bypassPermissions",
prompt: "..." # see references/spawn-prompt-guide.md
)
Every spawn prompt must be self-contained — teammates have zero conversation history. See references/spawn-prompt-guide.md for the full template and how to list each agent's available skills.
Messages from teammates arrive automatically. Check progress:
TaskList()
Answer a blocked teammate:
SendMessage(
type: "message",
recipient: "backend",
content: "Stripe webhook secret is STRIPE_WEBHOOK_SECRET in .env.local",
summary: "Stripe secret location"
)
SendMessage(type: "shutdown_request", recipient: "designer", content: "Work complete")
SendMessage(type: "shutdown_request", recipient: "backend", content: "Work complete")
SendMessage(type: "shutdown_request", recipient: "tester", content: "Work complete")
# Wait for each shutdown_response, then:
TeamDelete()
Parallel from the start:
├── researcher: research best practices / prior art
├── designer: UI components
├── nextjs or integration-expert: API / server logic
└── database: schema changes
Blocked until implementation complete:
├── tester: test suite
└── documentation-writer: feature docs
Parallel:
├── code-auditor: full vulnerability scan (Semgrep, CodeQL)
└── architecture-reviewer: structural/design issues
Blocked until audit complete:
├── nextjs or integration-expert: fix findings
└── tester: regression tests
Parallel:
├── code-auditor: security review
├── tester: coverage audit
├── optimizer: Lighthouse + bundle
├── documentation-writer: user-facing docs
└── legal: privacy / ToS
Blocked until all above complete:
└── devops: deploy pipeline
mode: "bypassPermissions" on every Agent spawn or teammates block| Problem | Fix |
|---|---|
TeamCreate fails | Check CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set |
| Teammate blocks/stalls | Missing mode: "bypassPermissions" — always set this |
| Teammate not claiming tasks | Check blockedBy deps with TaskGet |
| Teammate idle and unresponsive | Send a direct SendMessage — idle agents wake on receipt |
TeamDelete fails | Teammates still running. Send shutdown_request to each |
| Teammate went off-script | Send correction via SendMessage. If severe, shutdown and respawn |
references/permissions-and-isolation.md — mode parameter options, worktree isolation for parallel editsreferences/agent-roster.md — full roster table + which skills to mention per agent in spawn promptsreferences/spawn-prompt-guide.md — complete spawn prompt template with skills section