By PeterFile
Orchestrate multi-agent AI workflows for structured software engineering: generate specs (PRD/RFC), dispatch specialized agents (Codex, Gemini, Claude) for code tasks, enforce TDD, and sync documentation with code changes.
Execute codeagent-wrapper for multi-backend AI code tasks. Supports Codex, Claude, and Gemini backends with file references (@syntax) and structured output.
Execute Codex CLI for code analysis, refactoring, and automated code changes. Use when you need to delegate complex code tasks to Codex AI with file references (@syntax) and structured output.
Execute Gemini CLI for AI-powered code analysis and generation. Use when you need to leverage Google's Gemini models for complex reasoning tasks.
Create a Kiro spec for a feature in .kiro/specs create/update feature spec/PRD/RFC—requirements, design doc, and implementation tasks checklist.Trigger on spec/specification/PRD/RFC/tech spec, requirements/user story/acceptance criteria/EARS, design doc/architecture, task breakdown/implementation plan/checklist; 需求/验收/设计/任务.
Sync spec files with code changes. Triggers when modifying code that affects .kiro/specs/*/requirements.md or .kiro/specs/*/design.md. Use after implementing features, fixing bugs, or refactoring that changes behavior documented in specs.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimnpx claudepluginhub joshuarweaver/cascade-ai-ml-agents-misc-1 --plugin peterfile-devpilot-agentsBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
中文 | English
"亚瑟王的王者之剑已破石而出……你是否愿随王一同奔赴疆场?"
专为复杂软件工程任务设计的高保真 多代理编排系统。借鉴"圆桌会议"哲学,由 亚瑟王 (编排者) 协调专业代理,大规模地实现、评审并同步代码库变更。
设计哲学:通过 Spec-Driven Development 将需求拆解为原子任务,每个 Agent 只处理其上下文范围内的工作——无需理解全局,专注执行即可。任务分发给 Codex CLI (OpenAI) 和 Gemini CLI (Google) 等一线 Coding CLI,借力厂商的模型能力,而非重复造轮子去构建冗余的工具链和 sub-agent。
┌─────────────────────────────────────────────────────────────────┐
│ kiro-specs multi-agent-orchestrator │
│ ─────────── ──────────────────────── │
│ 需求 → 设计 → 任务 圆桌会议并行执行 tasks.md │
│ ↓ ↓ │
│ .kiro/specs/feature/tasks.md ──────→ AGENT_STATE.json │
└─────────────────────────────────────────────────────────────────┘
| 阶段 | 说明 |
|---|---|
| 1. 规范定义 | kiro-specs 引导创建 requirements → design → tasks |
| 2. 圆桌执行 | multi-agent-orchestrator 读取 tasks.md,分发给 Codex/Gemini (TDD) |
| 3. 评审闭环 | 每个任务经历 dispatch → review → fix → consolidate |
| 角色 | 代理 | 职责 |
|---|---|---|
| 亚瑟王 (King Arthur) | 编排者 | 规划、授权、质量门禁及项目状态同步。 |
| 高文 (Gawain) | 决策骑士 | 内环决策、严格的 JSON 输出及逻辑验证。 |
| Codex/Gemini | 执行者 | 分布式执行代码 (Codex) 与 UI (Gemini)。 |
| codeagent-wrapper | 执行层 | 驱动后端并行运行的 Go 运行时。 |
| 依赖项 | 版本 | 用途 |
|---|---|---|
| Node.js | 18+ | 运行 npx skills add |
| Python | 3.9+ | 编排脚本 |
| Go | 1.21+ | 编译 codeagent-wrapper |
| Claude Code/OpenCode | 最新版 | 触发技能 |
npx skills add PeterFile/devpilot-agents
此命令将本仓库的所有技能安装到你的代理环境中。
git clone https://github.com/PeterFile/devpilot-agents.git
cd devpilot-agents/codeagent-wrapper
go build -o codeagent-wrapper .
Windows:
go build -o codeagent-wrapper.exe .
Linux/macOS:
export PATH="$PWD:$PATH"
Windows (PowerShell):
$env:PATH = "$PWD;$env:PATH"
将代理定义复制到你的项目:
cp -r .opencode/agents/ <你的项目>/.opencode/agents/
或全局安装(对所有项目生效):
cp -r .opencode/agents/ ~/.config/opencode/agents/
这将添加 亚瑟王 (编排者) 和 高文 (决策骑士) 到你的 OpenCode 环境。
查看交互式流程图 - 点击逐步查看动画演示。
flowchart/ 目录包含源代码。本地运行:
cd flowchart
npm install
npm run dev
通过 codeagent-wrapper 调用 LLM 后端(默认 opencode),每轮迭代:
AGENT_STATE.json + PROJECT_PULSE.mddispatch_batch / dispatch_reviews 等动作| 变量 | 作用 | 默认值 |
|---|---|---|
CODEAGENT_OPENCODE_AGENT | 指定 OpenCode 使用的代理(可选) | OpenCode 默认 |
CODEAGENT_OPENCODE_MODEL | 覆盖 OpenCode 模型(可选) | OpenCode 配置 |
CODEAGENT_NO_TMUX | 禁用 tmux | Windows 自动设置 |
0:完成 | 1:停止/无进展 | 2:需人工决策在 Claude Code / OpenCode 中描述任务即可自动触发技能:
| 触发示例 | 技能 |
|---|---|
"Start orchestration from spec at .kiro/specs/my-feature" | multi-agent-orchestrator |
"Run orchestration for payment-integration" | multi-agent-orchestrator |
| "Create requirements for a new feature" | kiro-specs |
| "Help me write tests first" | test-driven-development |
| 技能 | 描述 |
|---|---|
| multi-agent-orchestrator | 从 Kiro 规范编排多代理工作流 |
| kiro-specs | 规范驱动工作流:需求 → 设计 → 任务 → 执行 |
| test-driven-development | 红-绿-重构 TDD 工作流 |
13 agents, 35 skills, 19 commands, 15 hooks — spec-driven, multi-agent orchestration for Claude Code with per-phase model tiers, opt-in lifecycle hooks, and optional cross-device semantic memory.
Universal multi-agent infrastructure: 26 specialist agents, 20 enforced workflow skills, and a lead orchestrator
This skill should be used when the model's ROLE_TYPE is orchestrator and needs to delegate tasks to specialist sub-agents. Provides scientific delegation framework ensuring world-building context (WHERE, WHAT, WHY) while preserving agent autonomy in implementation decisions (HOW). Use when planning task delegation, structuring sub-agent prompts, or coordinating multi-agent workflows.
Multi-agent development workflow (/crew) — planning, implementation, review, documentation, evals, and audit gates. Commands, agents, and skills are auto-discovered from the repo.
Persistent memory, shared standards, and structured workflows for AI coding agents. Detects project setup and injects agent context automatically.
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams