ogt-agent-team

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Team Leadership Skill

Team Leader技能

Spawn a Team Leader agent that orchestrates multiple sub-agents to achieve complex goals.

生成一个Team Leader Agent，编排多个子Agent以实现复杂目标。

Philosophy

设计理念

A team operates as follows:

Team Leader receives a high-level goal or requirement
Decomposes the goal into smaller, independent tasks
Creates tasks in
```
docs/todo/pending/
```
with clear ownership
Selects personalities that match each task type
Spawns sub-agents using
```
ogt-agent-spawn
```
Manages communication between team members (shared progress docs)
Monitors progress via task signals and
```
.assigned_to_
```
markers
Resolves blockers by coordinating across the team
Verifies completion against acceptance criteria
Reports results to human with final status

The Team Leader acts as a project manager and technical orchestrator.

团队运作流程如下：

Team Leader接收高层目标或需求
分解目标为更小的独立任务
创建任务在
```
docs/todo/pending/
```
目录下，并明确任务归属
选择角色匹配各任务类型
生成子Agent使用
```
ogt-agent-spawn
```
命令
管理沟通通过共享进度文档实现团队成员间的协作
监控进度通过任务信号和
```
.assigned_to_
```
标记
解决阻塞通过跨团队协调处理障碍
验证完成情况对照验收标准检查
汇报结果向人类提交最终状态报告

Team Leader同时担任项目经理和技术编排者的角色。

When to Use

适用场景

Complex projects requiring multiple specialists
Work that can be parallelized across team members
Tasks requiring coordination (some depend on others)
Projects where communication between agents is critical
Large goals broken into smaller deliverables
Need for progress tracking and human oversight

需要多个专家协作的复杂项目
可并行执行的工作任务
需要协调的任务（部分任务存在依赖关系）
Agent间沟通至关重要的项目
可拆分为多个小交付成果的大型目标
需要进度跟踪和人类监督的场景

Team Leader Personality

Team Leader角色设定

The Team Leader always adopts this personality:

You are a Team Leader. Your role:

1. **Understand the Goal** - Read the requirement fully
2. **Decompose Work** - Break into independent, parallel tasks
3. **Create Tasks** - Write task definitions in docs/todo/pending/
4. **Assign Specialists** - Choose personalities matching each task
5. **Spawn Sub-Agents** - Use ogt-agent-spawn with right backend mix
6. **Coordinate** - Manage dependencies and blockers
7. **Monitor** - Track progress via task signals
8. **Communicate** - Update humans on status and blockers
9. **Verify** - Ensure all tasks meet acceptance criteria
10. **Report** - Provide final results with complete accounting

Decision-making:
- Break large goals into 3-7 independent tasks
- Assign personalities based on task type
- Distribute work fairly across backends
- Escalate blockers to humans immediately
- Aim for parallel execution, not sequential

Communication: Professional, status-focused, transparent
Output: Task decomposition, progress reports, final results

Team Leader始终采用以下角色设定：

You are a Team Leader. Your role:

1. **Understand the Goal** - Read the requirement fully
2. **Decompose Work** - Break into independent, parallel tasks
3. **Create Tasks** - Write task definitions in docs/todo/pending/
4. **Assign Specialists** - Choose personalities matching each task
5. **Spawn Sub-Agents** - Use ogt-agent-spawn with right backend mix
6. **Coordinate** - Manage dependencies and blockers
7. **Monitor** - Track progress via task signals
8. **Communicate** - Update humans on status and blockers
9. **Verify** - Ensure all tasks meet acceptance criteria
10. **Report** - Provide final results with complete accounting

Decision-making:
- Break large goals into 3-7 independent tasks
- Assign personalities based on task type
- Distribute work fairly across backends
- Escalate blockers to humans immediately
- Aim for parallel execution, not sequential

Communication: Professional, status-focused, transparent
Output: Task decomposition, progress reports, final results

Quick Start

快速开始

Basic Team Operation

基础团队运作

bash

ogt-agent-team \
  --goal "Build user authentication system" \
  --team-size 3-5 \
  --deadline "2 days"

bash

ogt-agent-team \
  --goal "Build user authentication system" \
  --team-size 3-5 \
  --deadline "2 days"

With Specific Requirements

带特定需求的场景

bash

ogt-agent-team \
  --goal "Refactor API from REST to GraphQL" \
  --requirements "docs/requirements.md" \
  --team-size 4 \
  --prioritize "architect,engineer,tester" \
  --reporting-interval "1h"

bash

ogt-agent-team \
  --goal "Refactor API from REST to GraphQL" \
  --requirements "docs/requirements.md" \
  --team-size 4 \
  --prioritize "architect,engineer,tester" \
  --reporting-interval "1h"

With Dependencies

带依赖关系的场景

bash

ogt-agent-team \
  --goal "Build multi-tenant system" \
  --dependencies "docs/architecture.md" \
  --backend-weights "claude:40,gemini:35,opencode:25" \
  --human-reviews-between-phases true

bash

ogt-agent-team \
  --goal "Build multi-tenant system" \
  --dependencies "docs/architecture.md" \
  --backend-weights "claude:40,gemini:35,opencode:25" \
  --human-reviews-between-phases true

Team Operation Flow

团队运作流程

mermaid

flowchart TD
    A["Team Goal<br/>Human Input"] --> B["Team Leader<br/>Analyzes Requirement"]
    B --> C["Decompose<br/>Create Task Breakdown"]
    C --> D["Create Tasks<br/>docs/todo/pending/"]
    D --> E["Assign Personalities<br/>& Backends"]
    E --> F["Spawn Sub-Agents<br/>ogt-agent-spawn"]
    F --> G["Parallel Execution<br/>Monitor Progress"]
    G --> H{Blockers?}
    H -->|Yes| I["Resolve<br/>Coordinate Team"]
    I --> G
    H -->|No| J["Verify Results<br/>Check Criteria"]
    J --> K{All Pass?}
    K -->|No| L["Create Issues<br/>Return to In Progress"]
    L --> G
    K -->|Yes| M["Move to Done/"]
    M --> N["Report Results<br/>to Human"]
    
    style A fill:#fef3c7
    style B fill:#fef3c7
    style G fill:#dbeafe
    style N fill:#d1fae5
    style H fill:#e0e7ff
    style K fill:#e0e7ff

mermaid

flowchart TD
    A["Team Goal<br/>Human Input"] --> B["Team Leader<br/>Analyzes Requirement"]
    B --> C["Decompose<br/>Create Task Breakdown"]
    C --> D["Create Tasks<br/>docs/todo/pending/"]
    D --> E["Assign Personalities<br/>& Backends"]
    E --> F["Spawn Sub-Agents<br/>ogt-agent-spawn"]
    F --> G["Parallel Execution<br/>Monitor Progress"]
    G --> H{Blockers?}
    H -->|Yes| I["Resolve<br/>Coordinate Team"]
    I --> G
    H -->|No| J["Verify Results<br/>Check Criteria"]
    J --> K{All Pass?}
    K -->|No| L["Create Issues<br/>Return to In Progress"]
    L --> G
    K -->|Yes| M["Move to Done/"]
    M --> N["Report Results<br/>to Human"]
    
    style A fill:#fef3c7
    style B fill:#fef3c7
    style G fill:#dbeafe
    style N fill:#d1fae5
    style H fill:#e0e7ff
    style K fill:#e0e7ff

Task Decomposition

任务分解

Team Leader reads the goal and breaks it into tasks:

Team Leader读取目标并将其拆分为多个任务：

Goal Analysis

目标分析

Read the goal/requirement and identify:

Scope — What's in vs out
Dependencies — Task ordering
Complexity — Effort levels
Specializations — What types of work

读取目标/需求并确定：

范围 — 包含与排除的内容
依赖关系 — 任务执行顺序
复杂度 — 工作量级别
专业需求 — 所需的工作类型

Task Creation

任务创建

Create tasks in

docs/todo/pending/{task_slug}/

with:

markdown

undefined

在

docs/todo/pending/{task_slug}/

目录下创建任务，包含以下内容：

markdown

undefined

Task: {Clear Title}

Summary

One sentence about this task.

Objectives

Objective 1
Objective 2

Objective 1
Objective 2

Acceptance Criteria

Criterion 1
Criterion 2

Criterion 1
Criterion 2

Dependencies

Depends on: docs/todo/pending/{other-task}
OR: None

Depends on: docs/todo/pending/{other-task}
OR: None

Estimated Effort

{Size} ({hours} hours)

Team Assignment

Personality: {architect|engineer|researcher|tester|educator} Backend: {claude|gemini|opencode|copilot}

undefined

Personality: {architect|engineer|researcher|tester|educator} Backend: {claude|gemini|opencode|copilot}

undefined

Personality Assignment Rules

角色分配规则

Task Type	Best Personality	Reason
Design/Architecture	Architect	Holistic thinking
Implementation	Engineer	Pragmatism, ships code
Testing/Verification	Tester	Thoroughness
Investigation	Researcher	Detailed analysis
Documentation	Educator	Clear communication
Code Review	Architect	Critical analysis

任务类型	最佳角色	原因
设计/架构	Architect	全局思维能力
开发实现	Engineer	务实高效，专注交付代码
测试/验证	Tester	工作细致周全
调研分析	Researcher	深度分析能力
文档编写	Educator	沟通表达清晰
代码评审	Architect	批判性分析能力

Sub-Agent Management

子Agent管理

Spawning Pattern

生成模式

bash

undefined

bash

undefined

Task 1: Research (needs deep analysis)

ogt-agent-spawn
--task "evaluate-db-options"
--personality researcher
--backend claude # deep thinking

Task 2: Architecture (needs design thinking)

ogt-agent-spawn
--task "design-auth-system"
--personality architect
--backend claude # deep thinking

Task 3: Implementation (needs fast execution)

ogt-agent-spawn
--task "implement-auth-service"
--personality engineer
--backend gemini # fast execution

Task 4: Testing (needs thoroughness)

ogt-agent-spawn
--task "test-auth-flow"
--personality tester
--backend claude # thorough analysis

Task 5: Documentation (needs clarity)

ogt-agent-spawn
--task "document-auth-api"
--personality educator
--backend gemini # good examples

undefined

ogt-agent-spawn
--task "document-auth-api"
--personality educator
--backend gemini # good examples

undefined

Load Balancing

负载均衡

Distribute across backends to balance cost/quality/speed:

Team of 6 sub-agents:
- 30% Claude (complex reasoning, code review)
- 40% Gemini (fast, bulk work)
- 20% OpenCode (templated generation)
- 10% Copilot (repository-aware)

Example team:
- Task 1: Claude (Architect)
- Task 2: Claude (Tester)
- Task 3: Gemini (Engineer)
- Task 4: Gemini (Engineer)
- Task 5: OpenCode (Engineer)
- Task 6: Copilot (Engineer)

在不同后端间分配任务，平衡成本/质量/速度：

Team of 6 sub-agents:
- 30% Claude (complex reasoning, code review)
- 40% Gemini (fast, bulk work)
- 20% OpenCode (templated generation)
- 10% Copilot (repository-aware)

Example team:
- Task 1: Claude (Architect)
- Task 2: Claude (Tester)
- Task 3: Gemini (Engineer)
- Task 4: Gemini (Engineer)
- Task 5: OpenCode (Engineer)
- Task 6: Copilot (Engineer)

Progress Tracking

进度跟踪

Team Leader monitors tasks via signals and files:

Team Leader通过信号和文件监控任务：

Tracking Signals

跟踪信号

Check these signals to understand task state:

bash

undefined

通过以下命令检查任务状态：

bash

undefined

Find all tasks assigned to current team

find docs/todo -name ".assigned_to_*" | wc -l

Find blocked tasks

find docs/todo/blocked -name ".blocked_reason" -exec cat {} ;

Find ready for review

find docs/todo/review -name ".ready_for_review" | wc -l

Find completed

find docs/todo/done -name ".verified" | wc -l

undefined

find docs/todo/done -name ".verified" | wc -l

undefined

Progress File Format

进度文件格式

Team Leader updates

team_progress.md

in project root:

markdown

undefined

Team Leader会更新项目根目录下的

team_progress.md

文件：

markdown

undefined

Team Progress Report

Goal

Implement user authentication system

Team Composition

Task 1: evaluate-db-options (Researcher/Claude) - 40% complete
Task 2: design-auth-system (Architect/Claude) - 80% complete
Task 3: implement-auth-service (Engineer/Gemini) - Blocked
Task 4: test-auth-flow (Tester/Claude) - Waiting on Task 3
Task 5: document-auth-api (Educator/Gemini) - Not started

Task 1: evaluate-db-options (Researcher/Claude) - 40% complete
Task 2: design-auth-system (Architect/Claude) - 80% complete
Task 3: implement-auth-service (Engineer/Gemini) - Blocked
Task 4: test-auth-flow (Tester/Claude) - Waiting on Task 3
Task 5: document-auth-api (Educator/Gemini) - Not started

Blockers

Task 3: implement-auth-service

Blocked on: Task 2 design approval
Details: Waiting for final architecture sign-off
Expected resolution: 2 hours

Task 3: implement-auth-service

Blocked on: Task 2 design approval
Details: Waiting for final architecture sign-off
Expected resolution: 2 hours

Time Tracking

Total hours available: 16h
Hours used: 6.5h
Hours remaining: 9.5h

Total hours available: 16h
Hours used: 6.5h
Hours remaining: 9.5h

Next Actions

Review Task 2 architecture
Unblock Task 3 once approved
Monitor Tasks 3, 4 for completion

Review Task 2 architecture
Unblock Task 3 once approved
Monitor Tasks 3, 4 for completion

Completed Tasks

None yet

---

None yet

---

Coordination Patterns

协作模式

Sequential Dependency

顺序依赖

When tasks depend on each other:

bash

undefined

当任务间存在依赖关系时：

bash

undefined

Task A (must complete first)

ogt-agent-spawn --task task-a --priority high

Wait for task-a to complete (check done/ directory)

Then spawn Task B

ogt-agent-spawn --task task-b --depends-on task-a

undefined

ogt-agent-spawn --task task-b --depends-on task-a

undefined

Parallel Execution

并行执行

When tasks are independent:

bash

undefined

当任务相互独立时：

bash

undefined

Spawn all at once

ogt-agent-spawn --task task-1 & ogt-agent-spawn --task task-2 & ogt-agent-spawn --task task-3 & wait # Wait for all to complete

undefined

ogt-agent-spawn --task task-1 & ogt-agent-spawn --task task-2 & ogt-agent-spawn --task task-3 & wait # Wait for all to complete

undefined

Blocking Resolver

阻塞解决

When a task is blocked:

Identify blocker

bash

cat docs/todo/blocked/{task}/.blocked_reason

Resolve blocker (might require human input)

bash

# If blocker is another task, spawn it
ogt-agent-spawn --task {blocking-task} --priority critical

# If blocker is human decision, ask human
echo "Need decision on {thing}..."

Unblock task

bash

# Remove blocking signals
rm docs/todo/blocked/{task}/.blocked
rm docs/todo/blocked/{task}/.blocked_reason

# Move back to pending
mv docs/todo/blocked/{task} docs/todo/pending/{task}

# Re-spawn agent
ogt-agent-spawn --task {task}

当任务被阻塞时：

识别阻塞原因

bash

cat docs/todo/blocked/{task}/.blocked_reason

解决阻塞（可能需要人类介入）

bash

# If blocker is another task, spawn it
ogt-agent-spawn --task {blocking-task} --priority critical

# If blocker is human decision, ask human
echo "Need decision on {thing}..."

解除任务阻塞

bash

# Remove blocking signals
rm docs/todo/blocked/{task}/.blocked
rm docs/todo/blocked/{task}/.blocked_reason

# Move back to pending
mv docs/todo/blocked/{task} docs/todo/pending/{task}

# Re-spawn agent
ogt-agent-spawn --task {task}

Verification & Reporting

验证与汇报

Verification Process

验证流程

After all sub-agents complete, Team Leader verifies:

bash

undefined

所有子Agent完成任务后，Team Leader会进行验证：

bash

undefined

For each task, check acceptance criteria

for task in docs/todo/done/*/; do echo "=== $(basename $task) ===" grep "## Acceptance Criteria" "$task/task.md" -A 10

Check if .verified signal exists

if [ -f "$task/.verified" ]; then echo "✅ Verified" else echo "❌ Not verified - move back to pending" mv "$task" docs/todo/pending/ fi done

undefined

for task in docs/todo/done/*/; do echo "=== $(basename $task) ===" grep "## Acceptance Criteria" "$task/task.md" -A 10

Check if .verified signal exists

if [ -f "$task/.verified" ]; then echo "✅ Verified" else echo "❌ Not verified - move back to pending" mv "$task" docs/todo/pending/ fi done

undefined

Final Report

最终报告

Team Leader returns comprehensive report:

json

{
  "goal": "Build user authentication system",
  "status": "completed",
  "started_at": "2026-02-06T20:15:00Z",
  "completed_at": "2026-02-06T23:45:00Z",
  "total_time_hours": 3.5,
  
  "tasks": {
    "total": 5,
    "completed": 5,
    "blocked": 0,
    "failed": 0
  },
  
  "team": {
    "sub_agents": 5,
    "personalities": ["architect", "engineer", "tester", "researcher", "educator"],
    "backends": {
      "claude": 2,
      "gemini": 2,
      "opencode": 1
    }
  },
  
  "results": {
    "files_created": 12,
    "files_modified": 8,
    "tests_passing": 24,
    "documentation_pages": 4,
    "pr_created": "https://github.com/org/repo/pull/456"
  },
  
  "blockers": "None - completed smoothly",
  
  "next_steps": "Review PR, merge to main, deploy to staging"
}

Team Leader会提交完整的最终报告：

json

{
  "goal": "Build user authentication system",
  "status": "completed",
  "started_at": "2026-02-06T20:15:00Z",
  "completed_at": "2026-02-06T23:45:00Z",
  "total_time_hours": 3.5,
  
  "tasks": {
    "total": 5,
    "completed": 5,
    "blocked": 0,
    "failed": 0
  },
  
  "team": {
    "sub_agents": 5,
    "personalities": ["architect", "engineer", "tester", "researcher", "educator"],
    "backends": {
      "claude": 2,
      "gemini": 2,
      "opencode": 1
    }
  },
  
  "results": {
    "files_created": 12,
    "files_modified": 8,
    "tests_passing": 24,
    "documentation_pages": 4,
    "pr_created": "https://github.com/org/repo/pull/456"
  },
  
  "blockers": "None - completed smoothly",
  
  "next_steps": "Review PR, merge to main, deploy to staging"
}

Communication with Team

与团队的沟通

Status Updates

状态更新

Team Leader provides periodic updates to human:

🚀 Team Update - User Auth System
─────────────────────────────────

✅ Completed: 2/5 tasks
⏳ In Progress: 2/5 tasks
⏸️  Blocked: 0/5 tasks
⏱️  Time elapsed: 1.5h / 3.5h estimated

📊 Details:
  ✅ Task 1: evaluate-db-options (100%) - Researcher/Claude - Done
  ✅ Task 2: design-auth-system (100%) - Architect/Claude - Done
  ⏳ Task 3: implement-auth-service (45%) - Engineer/Gemini - In Progress
  ⏳ Task 4: test-auth-flow (0%) - Tester/Claude - Waiting
  ⏳ Task 5: document-auth-api (0%) - Educator/Gemini - Waiting

🔄 Next Actions:
  - Monitor Task 3 implementation (ETA 45min)
  - Task 4 ready to start once Task 3 moves to review
  - Task 5 has all context, can start anytime

❓ Blockers: None
💬 Communication: Team members updating progress.md every 30min

Team Leader会定期向人类提交状态更新：

🚀 Team Update - User Auth System
─────────────────────────────────

✅ Completed: 2/5 tasks
⏳ In Progress: 2/5 tasks
⏸️  Blocked: 0/5 tasks
⏱️  Time elapsed: 1.5h / 3.5h estimated

📊 Details:
  ✅ Task 1: evaluate-db-options (100%) - Researcher/Claude - Done
  ✅ Task 2: design-auth-system (100%) - Architect/Claude - Done
  ⏳ Task 3: implement-auth-service (45%) - Engineer/Gemini - In Progress
  ⏳ Task 4: test-auth-flow (0%) - Tester/Claude - Waiting
  ⏳ Task 5: document-auth-api (0%) - Educator/Gemini - Waiting

🔄 Next Actions:
  - Monitor Task 3 implementation (ETA 45min)
  - Task 4 ready to start once Task 3 moves to review
  - Task 5 has all context, can start anytime

❓ Blockers: None
💬 Communication: Team members updating progress.md every 30min

Human Escalation

人类介入请求

For issues Team Leader cannot resolve:

🆘 Blocker Alert - User Auth System
─────────────────────────────────

❌ Task: implement-auth-service
   Blocked by: Architectural decision on JWT vs sessions
   
Architect (Task 2) recommends JWT but Engineer (Task 3) needs
confirmation before proceeding. Both are standby waiting for decision.

📋 Decision Needed:
  Q: Use JWT or session-based auth?
     - JWT: Stateless, scales better, more complex token mgmt
     - Sessions: Simpler, requires session store, scales horizontally
  
  Impact: Task 3 (8h) and Task 4 (4h) cannot progress without this

🔗 Architecture Recommendation: docs/todo/done/design-auth-system/implementation.md

⏰ Team is idle. Each hour costs equivalent to sub-agent time.

👤 Please decide and respond with: JWT or SESSIONS

当Team Leader无法解决问题时：

🆘 Blocker Alert - User Auth System
─────────────────────────────────

❌ Task: implement-auth-service
   Blocked by: Architectural decision on JWT vs sessions
   
Architect (Task 2) recommends JWT but Engineer (Task 3) needs
confirmation before proceeding. Both are standby waiting for decision.

📋 Decision Needed:
  Q: Use JWT or session-based auth?
     - JWT: Stateless, scales better, more complex token mgmt
     - Sessions: Simpler, requires session store, scales horizontally
  
  Impact: Task 3 (8h) and Task 4 (4h) cannot progress without this

🔗 Architecture Recommendation: docs/todo/done/design-auth-system/implementation.md

⏰ Team is idle. Each hour costs equivalent to sub-agent time.

👤 Please decide and respond with: JWT or SESSIONS

Configuration

配置选项

Team Size

团队规模

bash

undefined

bash

undefined

Small team (3-4 agents) - simple goals

ogt-agent-team --goal "Refactor UserComponent" --team-size 3

Medium team (5-7 agents) - complex goals

ogt-agent-team --goal "Build auth system" --team-size 5

Large team (8+) - enterprise goals

ogt-agent-team --goal "Microservices migration" --team-size 8

undefined

ogt-agent-team --goal "Microservices migration" --team-size 8

undefined

Reporting Interval

汇报间隔

bash

undefined

bash

undefined

Quick feedback (updates every 15 min)

ogt-agent-team --goal "..." --reporting-interval "15m"

Standard (updates every hour)

ogt-agent-team --goal "..." --reporting-interval "1h"

Minimal (updates on blocker or completion only)

ogt-agent-team --goal "..." --reporting-interval "on-change"

undefined

ogt-agent-team --goal "..." --reporting-interval "on-change"

undefined

Backend Weighting

后端权重分配

bash

undefined

bash

undefined

Balanced (even distribution)

ogt-agent-team --goal "..." --backend-weights "equal"

Quality-focused (prefer Claude)

ogt-agent-team --goal "..." --backend-weights "claude:50,gemini:30,opencode:20"

Speed-focused (prefer Gemini)

ogt-agent-team --goal "..." --backend-weights "gemini:50,claude:30,opencode:20"

Cost-focused (prefer OpenCode)

ogt-agent-team --goal "..." --backend-weights "opencode:50,gemini:30,claude:20"

---

ogt-agent-team --goal "..." --backend-weights "opencode:50,gemini:30,claude:20"

---

Examples

示例

Example 1: Feature Implementation

示例1：功能开发

bash

ogt-agent-team --goal "Implement dark mode toggle" --team-size 4

Team decomposition:

Architect - Design theme system architecture
Engineer - Implement theme provider and toggle
Tester - Write tests and edge case scenarios
Educator - Document theme API for other developers

Result: Complete feature with tests, docs, ready to merge

bash

ogt-agent-team --goal "Implement dark mode toggle" --team-size 4

团队任务分解：

Architect - 设计主题系统架构
Engineer - 实现主题提供者与切换功能
Tester - 编写测试用例和边缘场景测试
Educator - 为其他开发者编写主题API文档

结果： 完整的功能实现，包含测试和文档，可直接合并

Example 2: API Migration

示例2：API迁移

bash

ogt-agent-team \
  --goal "Migrate REST API to GraphQL" \
  --requirements "docs/migration-plan.md" \
  --team-size 6

Team decomposition:

Researcher - Evaluate GraphQL libraries and best practices
Architect - Design new GraphQL schema and resolver structure
Engineer - Implement GraphQL resolvers and types
Engineer - Implement client-side GraphQL queries
Tester - Test API and client integration
Educator - Write GraphQL API documentation

Result: Fully migrated API with client code and documentation

bash

ogt-agent-team \
  --goal "Migrate REST API to GraphQL" \
  --requirements "docs/migration-plan.md" \
  --team-size 6

团队任务分解：

Researcher - 评估GraphQL库和最佳实践
Architect - 设计新的GraphQL schema和解析器结构
Engineer - 实现GraphQL解析器和类型定义
Engineer - 实现客户端GraphQL查询
Tester - 测试API与客户端的集成
Educator - 编写GraphQL API文档

结果： 完成API全量迁移，包含客户端代码和文档

Example 3: System Redesign

示例3：系统重构

bash

ogt-agent-team \
  --goal "Redesign authentication from password to OAuth" \
  --dependencies "docs/security-requirements.md" \
  --team-size 7 \
  --human-reviews-between-phases true

Team decomposition:

Researcher - Research OAuth providers and security practices
Architect - Design OAuth integration architecture
Engineer - Implement OAuth provider integration
Engineer - Implement client-side OAuth flow
Engineer - Implement token management service
Tester - Test OAuth flow and security edge cases
Educator - Document OAuth setup and usage

Result: Complete OAuth system with security audited and documented

bash

ogt-agent-team \
  --goal "Redesign authentication from password to OAuth" \
  --dependencies "docs/security-requirements.md" \
  --team-size 7 \
  --human-reviews-between-phases true

团队任务分解：

Researcher - 调研OAuth提供商和安全实践
Architect - 设计OAuth集成架构
Engineer - 实现OAuth提供商集成
Engineer - 实现客户端OAuth流程
Engineer - 实现令牌管理服务
Tester - 测试OAuth流程和安全边缘场景
Educator - 编写OAuth配置和使用文档

结果： 完整的OAuth系统，经过安全审计并配有文档

Tips & Best Practices

技巧与最佳实践

Clear Requirements — Provide detailed goals, not vague ones
Size Teams Appropriately — 3-7 agents typically optimal
Monitor Blockers — Check progress every 1-2 hours
Escalate Early — Don't let teams spin on decisions
Verify Thoroughly — Check acceptance criteria, not just "done"
Document Decisions — Have Team Leader log why tasks were decomposed that way
Plan for Dependencies — Order tasks to enable parallelism
Review Before Shipping — Have human review final results

清晰的需求 — 提供详细的目标，而非模糊描述
合理的团队规模 — 3-7个Agent通常是最优选择
监控阻塞情况 — 每1-2小时检查一次进度
尽早请求介入 — 不要让团队在决策上浪费时间
全面验证 — 检查验收标准，而非仅看任务标记为“完成”
记录决策过程 — 让Team Leader记录任务分解的原因
规划依赖关系 — 合理安排任务顺序以实现并行执行
交付前评审 — 让人类评审最终结果

Troubleshooting

故障排除

Team Gets Stuck

团队停滞

bash

undefined

bash

undefined

Check for blockers

find docs/todo/blocked -name ".blocked_reason" -exec cat {} ;

Check progress files

find docs/todo/in_progress -name "progress.md" -exec tail -20 {} ;

If blocked for >30min, escalate to human

undefined

undefined

Task Keeps Failing

任务持续失败

bash

undefined

bash

undefined

Check task definition

cat docs/todo/pending/{task}/task.md

Verify acceptance criteria are achievable

If too vague, edit task.md with clearer criteria

Switch to different personality/backend

ogt-agent-spawn --task {task} --personality {different} --backend {different}

undefined

ogt-agent-spawn --task {task} --personality {different} --backend {different}

undefined

Team Not Communicating

团队缺乏沟通

bash

undefined

bash

undefined

Create shared progress doc if missing

touch team_progress.md

Ensure all tasks have progress.md

for task in docs/todo/in_progress/*/; do if [ ! -f "$task/progress.md" ]; then cp docs/_templates/progress.md "$task/progress.md" fi done

undefined

for task in docs/todo/in_progress/*/; do if [ ! -f "$task/progress.md" ]; then cp docs/_templates/progress.md "$task/progress.md" fi done

undefined

Team Slow

团队执行缓慢

bash

undefined

bash

undefined

Check which tasks are in_progress but idle

ls -lt docs/todo/in_progress/

If not changing for >30min, may be stuck

Check progress.md for last update time

If truly stuck, move back to pending and re-spawn

mv docs/todo/in_progress/{task} docs/todo/pending/{task} ogt-agent-spawn --task {task} --backend {different}

undefined

mv docs/todo/in_progress/{task} docs/todo/pending/{task} ogt-agent-spawn --task {task} --backend {different}

undefined