{"_id":"@deep-flux/liubu","_rev":"5-4f98522e3ace2d43a19f647cbb504063","name":"@deep-flux/liubu","dist-tags":{"latest":"3.1.2"},"versions":{"3.0.1":{"name":"@deep-flux/liubu","version":"3.0.1","keywords":["opencode","plugin","multi-agent","workflow","三省六部制","hierarchical-governance","ai-orchestration"],"author":{"name":"Sansheng Liubu Contributors"},"license":"MIT","_id":"@deep-flux/liubu@3.0.1","maintainers":[{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"}],"homepage":"https://github.com/yourusername/sansheng-liubu#readme","bugs":{"url":"https://github.com/yourusername/sansheng-liubu/issues"},"bin":{"liubu":"bin/cli.cjs"},"dist":{"shasum":"ffe9ffa9f1f855f2b6194d36c5da16987aa250d9","tarball":"https://registry.npmjs.org/@deep-flux/liubu/-/liubu-3.0.1.tgz","fileCount":394,"integrity":"sha512-m6u+Qiw1S+METIJ2KY+xYOX51ltuRJkld9cL7yP/LY5BtYI4YpBf/RMFbPxyAed62SdqLc041UMR5/Ml6RgAZQ==","signatures":[{"sig":"MEYCIQDM28otTDScy4izf6oapwrqws/sRDUuoN4RyOULj1DFRgIhAIlcKM1x9YwOgOPp6chz8Sq47ul7QoE7YodPaCNbjBCJ","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":1119688},"main":"dist/index.js","type":"module","types":"dist/index.d.ts","engines":{"node":">=16.0.0"},"gitHead":"26f8e686538791fbfdac0daae9eba4d0214c65b0","scripts":{"test":"jest","build":"tsc","clean":"rm -rf dist","test:watch":"jest --watch","postinstall":"node bin/cli.cjs install","test:coverage":"jest --coverage","prepublishOnly":"npm run build"},"_npmUser":{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"},"repository":{"url":"git+https://github.com/yourusername/sansheng-liubu.git","type":"git"},"_npmVersion":"10.9.2","description":"三省六部制 OpenCode Plugin - 完整的多智能体协作框架，包含皇帝（战略决策）、三省（规划审核执行）、六部（具体实现）的分层架构","directories":{},"_nodeVersion":"23.11.1","_hasShrinkwrap":false,"devDependencies":{"jest":"^29.7.0","ts-jest":"^29.1.1","typescript":"^5.9.3","@types/jest":"^29.5.11","@types/node":"^20.19.37","@opencode-ai/sdk":"1.2.26","@opencode-ai/plugin":"latest"},"peerDependencies":{"@opencode-ai/sdk":"*","@opencode-ai/plugin":"*"},"_npmOperationalInternal":{"tmp":"tmp/liubu_3.0.1_1773670654896_0.33664597808004393","host":"s3://npm-registry-packages-npm-production"}},"3.1.1":{"name":"@deep-flux/liubu","version":"3.1.1","keywords":["opencode","plugin","multi-agent","workflow","三省六部制","hierarchical-governance","ai-orchestration"],"author":{"name":"Sansheng Liubu Contributors"},"license":"MIT","_id":"@deep-flux/liubu@3.1.1","maintainers":[{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"}],"homepage":"https://github.com/yourusername/sansheng-liubu#readme","bugs":{"url":"https://github.com/yourusername/sansheng-liubu/issues"},"bin":{"liubu":"bin/cli.cjs"},"dist":{"shasum":"b9fc29c311e4694d17e75618328b7269ddb67f75","tarball":"https://registry.npmjs.org/@deep-flux/liubu/-/liubu-3.1.1.tgz","fileCount":414,"integrity":"sha512-AHVVtdJfmd9JxAb1mk3sM+nTj+8Pm2XnJ/X9U+Iqpdcg9l3ZAAc+wXYJ+JvyhrFPtyVHAtE2hV3OKdy6xVbD7A==","signatures":[{"sig":"MEQCIGzJhCGiuED06Fq04dYSDWpzvaTKz+55Q0AfKsBabFo+AiBLYJYsZ2EYGj9hDBpqLKQqBt3pzXVrCW1Rj2XIAwWr+g==","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":1227965},"main":"dist/index.js","type":"module","types":"dist/index.d.ts","engines":{"node":">=16.0.0"},"gitHead":"32f8e665b5fa2252de29d9430b610003b18ea610","scripts":{"test":"jest","build":"tsc","clean":"rm -rf dist","test:watch":"jest --watch","postinstall":"node bin/cli.cjs install","test:coverage":"jest --coverage","prepublishOnly":"npm run build"},"_npmUser":{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"},"repository":{"url":"git+https://github.com/yourusername/sansheng-liubu.git","type":"git"},"_npmVersion":"10.9.2","description":"三省六部制 OpenCode Plugin - 完整的多智能体协作框架，包含皇帝（战略决策）、三省（规划审核执行）、六部（具体实现）的分层架构","directories":{},"_nodeVersion":"23.11.1","_hasShrinkwrap":false,"devDependencies":{"jest":"^29.7.0","ts-jest":"^29.1.1","typescript":"^5.9.3","@types/jest":"^29.5.11","@types/node":"^20.19.37","@opencode-ai/sdk":"1.2.26","@opencode-ai/plugin":"latest"},"peerDependencies":{"@opencode-ai/sdk":"*","@opencode-ai/plugin":"*"},"_npmOperationalInternal":{"tmp":"tmp/liubu_3.1.1_1773682544332_0.9175225863359966","host":"s3://npm-registry-packages-npm-production"}},"3.1.2":{"name":"@deep-flux/liubu","version":"3.1.2","type":"module","description":"三省六部制 OpenCode Plugin - 完整的多智能体协作框架，包含皇帝（战略决策）、三省（规划审核执行）、六部（具体实现）的分层架构","main":"dist/index.js","types":"dist/index.d.ts","keywords":["opencode","plugin","multi-agent","workflow","三省六部制","hierarchical-governance","ai-orchestration"],"author":{"name":"Sansheng Liubu Contributors"},"license":"MIT","repository":{"type":"git","url":"git+https://github.com/er-three/sansheng.git"},"publishConfig":{"registry":"https://registry.npmjs.org/"},"bin":{"liubu":"bin/cli.cjs"},"scripts":{"build":"tsc","test":"jest","test:watch":"jest --watch","test:coverage":"jest --coverage","postinstall":"node bin/cli.cjs install","prepublishOnly":"npm run build","clean":"rm -rf dist"},"engines":{"node":">=16.0.0"},"peerDependencies":{"@opencode-ai/plugin":"*","@opencode-ai/sdk":"*"},"devDependencies":{"@opencode-ai/plugin":"latest","@opencode-ai/sdk":"1.2.26","@types/jest":"^29.5.11","@types/node":"^20.19.37","jest":"^29.7.0","ts-jest":"^29.1.1","typescript":"^5.9.3"},"_id":"@deep-flux/liubu@3.1.2","gitHead":"084ad85a8a1fe856b4ad945f08b4789b2c2053c1","bugs":{"url":"https://github.com/er-three/sansheng/issues"},"homepage":"https://github.com/er-three/sansheng#readme","_nodeVersion":"23.11.1","_npmVersion":"10.9.2","dist":{"integrity":"sha512-4QI69TWruo2UynNe7lv01umeyIEMrabko1mtW/Y1elYYHMmrzO1Pg6LX0sjGEfXeMsJetRiHEJoCLYLlFCxTjA==","shasum":"a0d423b64be8f0c74efeeb4ee4396aace466c2d7","tarball":"https://registry.npmjs.org/@deep-flux/liubu/-/liubu-3.1.2.tgz","fileCount":415,"unpackedSize":1230310,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEQCIC6wH9iGSYglbdxFDcshL6x+zwh+NP/EBi/Ebz5sKrESAiAFH7wzIEM95Zu3/b+m3pfuAeuS4kYzNtGS1nhyierQug=="}]},"_npmUser":{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"},"directories":{},"maintainers":[{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/liubu_3.1.2_1773768951133_0.36825367314663593"},"_hasShrinkwrap":false}},"time":{"created":"2026-03-16T14:17:34.779Z","modified":"2026-03-17T17:35:51.405Z","1.0.0-beta.1":"2026-03-14T13:25:28.392Z","3.0.1":"2026-03-16T14:17:35.070Z","3.1.1":"2026-03-16T17:35:44.539Z","3.1.2":"2026-03-17T17:35:51.293Z"},"bugs":{"url":"https://github.com/er-three/sansheng/issues"},"author":{"name":"Sansheng Liubu Contributors"},"license":"MIT","homepage":"https://github.com/er-three/sansheng#readme","keywords":["opencode","plugin","multi-agent","workflow","三省六部制","hierarchical-governance","ai-orchestration"],"repository":{"type":"git","url":"git+https://github.com/er-three/sansheng.git"},"description":"三省六部制 OpenCode Plugin - 完整的多智能体协作框架，包含皇帝（战略决策）、三省（规划审核执行）、六部（具体实现）的分层架构","maintainers":[{"name":"deep-flux","email":"jhon.mobility.app@gmail.com"}],"readme":"# @deep-flux/liubu 3.1.1\n\n企业级多智能体协作框架，基于三省六部制设计。\n\n**最新版本**: 3.1.1 (2026-03-17) - [OK] 权限矩阵强制执行 + task()调用方式修正\n\n---\n\n## 核心特点\n\n- **11个智能体** - 皇帝战略决策、三省规划审核执行、六部具体实现\n- **4层验证网关** - 代码修改前自动检查：工作流状态→风险评估→审核验证→执行判定\n- **持久化审计** - 所有操作完整追踪到文件（`.opencode/audit/{sessionId}.json`）\n- **多层审核** - 皇帝→中书省→门下省→尚书省→六部的严格把关流程\n- **Task依赖** - 任务队列管理，自动检查前置条件完成\n- **测试强制** - 失败自动阻塞，防止缺陷继续扩散\n- **Agent心跳** - 实时监控任务状态，超时自动预警\n\n---\n\n## 优势\n\n| 优势 | 说明 |\n|------|------|\n| **自动化** | 零配置自动发现，开箱即用 |\n| **安全性** | 4层前置拦截，多层审核把关，完整审计追踪 |\n| **可靠性** | Task依赖检查，测试强制，自动重试 |\n| **可观测** | Agent心跳、完整日志、诊断工具 |\n| **易维护** | 清晰的层级结构，职责分工明确 |\n| **可扩展** | Recipe系统编排，Skill灵活组合 |\n\n---\n\n## 快速使用\n\n### 安装\n\n```bash\nnpm install @deep-flux/liubu\n```\n\n### 配置\n\n在 `opencode.json` 中添加：\n\n```json\n{\n  \"plugin\": [\"@deep-flux/liubu\"],\n  \"default_agent\": \"huangdi\"\n}\n```\n\n### 启动任务\n\n```bash\n# 方式1：通用编程任务\n/start 实现用户登录功能\n\n# 方式2：变更请求处理\n/start 升级API到v2版本，添加OAuth认证\n\n# 方式3：查看状态\n/status\n```\n\n### 工作流程\n\n```\n用户输入(/start)\n  [down]\n皇帝分析意图 → 选择工作域 → 分配任务给三省\n  [down]\n中书省制定计划\n  [down]\n门下省审核检查\n  [down]\n尚书省执行协调 → 调用六部完成\n  [down]\n结果验收\n```\n\n---\n\n## 工作域支持\n\n框架支持两个核心工作域：\n\n### [1] General - 通用编程\n用于标准的编程任务、bug修复和功能开发\n- **quick_fix** - 快速修复 (5分钟, ~70K token)\n- **standard** - 标准流程 (10分钟, ~120K token)\n- **comprehensive** - 强化审计 (15分钟, ~180K token)\n\n### [2] CR Processing - 变更请求处理\n用于系统升级、API变更和版本更新\n- **hotfix** - 紧急修复 (5分钟, ~90K token)\n- **standard** - 标准CR流程 (12分钟, ~140K token)\n- **complete** - 完整版本管理 (20分钟, ~200K token)\n\n---\n\n## 工具支持和调用\n\n### 代码修改工具 (自动验证)\n\n这些工具的调用会自动触发4层网关验证：\n\n```typescript\n// Edit - 修改文件\nEdit(file_path, old_string, new_string)\n\n// Write - 创建/覆盖文件\nWrite(file_path, content)\n\n// NotebookEdit - 修改Jupyter Notebook\nNotebookEdit(notebook_path, cell_number, new_source)\n```\n\n**验证流程**：\n1. 工作流初始化检查\n2. 风险评估（涉及文件数、行数、文件类型）\n3. 审核需求判断\n4. 最终执行判定\n\n---\n\n### 任务管理工具\n\n```typescript\n// 声明开始任务\n@claim_task task_id\n\n// 标记任务完成\n@complete_task task_id\n\n// 声明测试结果（失败时阻塞后续修改）\n@declare_test_result task_id passed|failed \"description\"\n\n// 查看任务队列\n@get_task_queue\n\n// 查看任务状态\n@task_status task_id\n```\n\n---\n\n### Skill调用\n\n```typescript\n// 通过 skill 调用执行具体功能\ntask(agent=\"agent_name\", skill=\"skill_name\", prompt=\"参数\")\n\n// 例子\ntask(agent=\"gongbu\", skill=\"code_implement\", prompt=\"实现登录功能\")\ntask(agent=\"bingbu\", skill=\"run_tests\", prompt=\"运行单元测试\")\ntask(agent=\"xingbu\", skill=\"code_review\", prompt=\"审查修改的代码\")\n```\n\n---\n\n### 日志调用\n\n```typescript\n// 标准日志输出（自动发送到OpenCode控制台）\nlog(\"ComponentName\", \"message\", level)\n\n// 诊断日志系统\ndiagnoseLoggerStatus()  // 返回：ready | degraded | unavailable\n```\n\n---\n\n## 系统架构\n\n### 智能体层级\n\n```\n【战略层】\n皇帝 - 接收意图、分配任务、最终验收\n\n【执行层 - 三省】\n├─ 中书省 - 制定计划\n├─ 门下省 - 审核验证\n└─ 尚书省 - 执行协调\n\n【工作层 - 六部】\n├─ 吏部 - 档案采集\n├─ 户部 - 资源整合\n├─ 礼部 - 工作流协调\n├─ 兵部 - 测试执行\n├─ 刑部 - 代码审查\n└─ 工部 - 代码实现\n```\n\n### 权限矩阵\n\n| Agent | 可调用SubAgent | 权限 |\n|-------|--------|------|\n| 皇帝 | 中书省、门下省、尚书省 | 战略决策 |\n| 中书省 | 门下省 | 规划制定 |\n| 门下省 | 无 | 审核验证 |\n| 尚书省 | task()调用六部 | 执行协调 |\n| 六部 | 无 | 具体实现 |\n\n---\n\n## 扩展方案\n\n### 1. 添加新的Skill\n\n在 `.opencode/domains/{domain}/skills/` 下创建：\n\n```yaml\n# skill-name.md\n---\ndescription: 技能描述\nmode: subagent\nmodel: anthropic/claude-haiku-4-5\ntemperature: 0.0\n---\n\n你是{Agent}，职责是{具体职责}\n\n## 输入参数\n\n- param1: 说明\n\n## 输出格式\n\n```json\n{\n  \"status\": \"success|failed\",\n  \"result\": \"具体结果\"\n}\n```\n```\n\n### 2. 添加新的Recipe\n\n在 `.opencode/recipes/` 下创建YAML文件：\n\n```yaml\nname: 工作流名称\ntype: simple|complex\ncomplexity: simple|medium|complex\nsteps:\n  - id: step-1\n    name: 第一步\n    description: 执行说明\n    uses: [agent1, agent2]\n    skill: skill_name\n    required_inputs:\n      - input1\n      - input2\n  - id: step-2\n    name: 第二步\n    depends_on: [step-1]\n    uses: [agent3]\n    skill: another_skill\n```\n\n### 3. 添加新的Agent\n\n在 `.opencode/agents/` 下创建：\n\n```yaml\n---\ndescription: Agent职责描述\nmode: subagent\nmodel: anthropic/claude-haiku-4-5\ntemperature: 0.0\npermission:\n  task:\n    agent_name: allow\n  skill:\n    \"*\": allow\nallowed_tools:\n  - task\n  - skill\n---\n\n你是{新Agent}，职责是：\n- 职责1\n- 职责2\n\n## 工作流程\n\n1. 接收任务\n2. 执行逻辑\n3. 返回结果\n```\n\n### 4. 自定义约束\n\n在 `.opencode/constraints/` 下创建：\n\n```markdown\n## 约束名称\n\n- 约束内容1\n- 约束内容2\n```\n\n自动加载，无需配置。\n\n---\n\n## 性能指标\n\n| 操作 | 延迟 | 说明 |\n|------|------|------|\n| 代码修改网关 | <1ms | 4层验证 |\n| 审计记录追加 | <10ms | 文件I/O |\n| 测试状态检查 | <1ms | 内存查询 |\n| 日志输出 | <5ms | OpenCode API |\n\n---\n\n## 常见问题\n\n**Q: 代码修改被拒绝？**\nA: 检查：工作流初始化 → 任务声明 → 依赖完成 → 审核状态\n\n**Q: 如何追踪修改历史？**\nA: 查看 `.opencode/audit/{sessionId}.json`\n\n**Q: 如何添加新的验证规则？**\nA: 修改 `src/workflows/code-modification-gateway.ts`\n\n**Q: 如何自定义Agent职责？**\nA: 在 `.opencode/agents/{agent_name}.md` 中更新 `permission` 和 `allowed_tools`\n\n**Q: 测试失败后如何继续？**\nA: 修复失败原因，调用 `@declare_test_result task_id passed`\n\n---\n\n## 文档\n\n所有核心文档位于 `docs/` 文件夹：\n\n- **docs/index.md** - 📚 文档导航索引\n- **docs/quick-start.md** - 🚀 5分钟快速开始\n- **docs/agents.md** - 🤖 11个智能体详解\n- **docs/architecture.md** - [ARCHITECTURE] 系统架构设计\n- **docs/phase-3.md** - ⚙️ Phase 3核心功能（网关、审计、测试强制）\n- **docs/opencode-logger.md** - [note] 日志系统说明\n- **docs/plugin-principles.md** - 🔧 系统设计原则\n\n---\n\n## 许可证\n\nMIT\n\n---\n\n**当前版本**: 3.0.1 | **发布日期**: 2026-03-16 | **状态**: [OK] 生产就绪\n\n### 版本3.0.1更新\n- 删除资产整理功能\n- 删除逆向代码功能\n- 删除Agent流程优化模块\n- 专注于核心：通用编程和变更请求处理\n- 代码清理：删除1732行不必要代码\n\n用 `/start` 开始第一个任务！\n","readmeFilename":"README.md"}