orchestrate

name	orchestrate
description	Agent orchestrator. Analyzes tasks and launches 2-4 specialized agents in parallel, then synthesizes results. 12 agents built from analysis of real usage patterns. Use when: (1) user says "/orchestrate", "run orchestrator", (2) task benefits from parallel analysis by multiple specialists, (3) complex task needs research + implementation or analysis + verification simultaneously.

Orchestrate

Launches specialized agents in parallel for complex tasks.

Workflow

1. Load library from agents-library.json (located in the same directory as this skill)
2. Select agents by task keywords (max 4)
3. Launch in parallel via Task tool
4. Synthesize results into a final report

---

Task to Agent Mapping

Keywords	Agents
research, find, information	researcher, triangulator
write code, implement, function, component	developer, tester
bug, error, fix, not working	debugger, developer
deploy, vercel, railway, push, git	devops
analyze, evaluate, review, weak spots	analyst, cto
design, ui, frontend, style	designer
text, post, article, documentation, readme	writer
plan, strategy, roadmap, priorities	strategist
ultrathink, deep, detailed	ultrathink, analyst
triangulate, verify fact, verification	triangulator, researcher
test, validation	tester, developer

---

12 Agents

Detailed system prompts are in agents-library.json.

ID	Name	Specialization
researcher	Deep Researcher	Web search, 3+ sources, verification
developer	Code Developer	Code, TypeScript/Python/React
debugger	Systematic Debugger	Root cause, systematic debugging
devops	DevOps Engineer	Deploy, CI/CD, Git, Vercel, Railway
analyst	Business Analyst	Business analysis, recommendations
designer	Frontend Designer	UI/UX, React, Tailwind, shadcn
writer	Content Writer	Text, posts, documentation
strategist	Product Strategist	Planning, roadmap
cto	CTO Advisor	Code review, architecture, security
triangulator	Fact Triangulator	Fact verification through 3+ sources
tester	QA Engineer	TDD, pytest, Jest, edge cases
ultrathink	Deep Thinker	Deep analysis, contrarian view

---

Instructions

On receiving a task with /orchestrate:

Step 1: Load library

Read {skill_directory}/agents-library.json

Step 2: Select agents

Pick 2-4 agents using the mapping above. Prefer agents with higher priority.

Step 3: Launch in parallel

For each agent, launch a Task tool simultaneously:

Task tool:
- description: "{agent.name} - {brief task description}"
- subagent_type: "general-purpose"
- prompt: "{agent.system_prompt}\n\n---\n\nTASK: {user_task}\n\n---\n\nComplete the task according to your specialization.\nFormat: {agent.output_format}"

CRITICAL: All Task tool calls in ONE message for parallel execution.

Step 4: Synthesis

After receiving all results:

1. Merge key findings
2. Resolve contradictions (flag if present)
3. Produce a structured report:
   • Executive Summary
   • Results by agent (brief)
   • Synthesized conclusions
   • Prioritized recommendations

---

Examples

Business analysis

/orchestrate Analyze the business model of company X and find weak spots

Agents: analyst, researcher, strategist, cto

Debugging

/orchestrate Bug: API returns 500 on large payload

Agents: debugger, developer

Research

/orchestrate Find best practices for RAG in 2026

Agents: researcher, triangulator

Deep analysis

/orchestrate ultrathink: why does retention drop after onboarding

Agents: ultrathink, analyst, researcher, strategist

Content creation

/orchestrate Write a landing page for an AI course with modern design

Agents: designer, writer, developer

---

Extension

To add an agent, edit agents-library.json:

{
  "id": "new_agent",
  "name": "Agent Name",
  "system_prompt": "...",
  "typical_tasks": ["task 1", "task 2"],
  "tools": ["Read", "Write"],
  "output_format": "structured_report"
}

Then add the mapping in orchestration_rules.task_to_agents_mapping.