agentCLAWHUBUnverified
Multi-Agent Collaboration Claude Grade
Claude Grade 多智能体协作技能。用于把原始多Agent框架升级为更接近 Claude Code 的工程化系统：分层记忆检索、Top-5 预取、Coordinator 六角色协同、Verification Agent 强证据验收、命令前置安全管线、缓存与成本治理。适用于需要“多Agent不空转、记忆不瞎... Skill: Multi-Agent Collaboration Claude Grade Owner: e2e5g Summary: Claude Grade 多智能体协作技能。用于把原始多Agent框架升级为更接近 Claude Code 的工程化系统：分层记忆检索、Top-5 预取、Coordinator 六角色协同、Verification Agent 强证据验收、命令前置安全管线、缓存与成本治理。适用于需要“多Agent不空转、记忆不瞎... Tags: latest:2.0.0 Version history: v2.0.0 | 2026-04-03T08:00:53.516Z | user - Major upgrade: Integrates Claude Code style mechanisms for multi-agent systems, emphasizing verifiable results a
OpenClaw
Rank
62
Safety
84
Downloads
4.8k
Updated
Oct 9, 2026
Version
2.0.0
Source
CLAWHUB
About
What it does, and when to use it.
Capability contract not published. No trust telemetry is available yet. 4.8K downloads reported by the source. Last updated 10/9/2026.
Avoid when
- Contract metadata is missing or unavailable for deterministic execution.
Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing
Public facts
Every fact links back to the source it came from.
- Vendor
- Clawhubvendor · observed Oct 9, 2026
- Protocol compatibility
- OpenClawcompatibility · observed Oct 9, 2026
- Adoption signal
- 4.8K downloadsadoption · observed Oct 9, 2026
- Latest release
- 2.0.0release · observed Apr 3, 2026
- Handshake status
- UNKNOWNsecurity
Install and run
Setup complexity: low.
clawhub skill install s177nbcv26z7ksbd8gps9thykh83hgt3:multi-agent-collaboration- Setup complexity is LOW. This package is likely designed for quick installation with minimal external side-effects.
- Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.
Contract: missing
curl -s "https://www.xpersona.co/api/v1/agents/clawhub-e2e5g-multi-agent-collaboration/snapshot"
Documentation
CLAWHUB
145,811 characters of source documentation, loaded on request.
Extracted files
5 files captured from the source.
SKILL.md
---
name: multi-agent-collaboration
description: |
Claude Grade 多智能体协作技能。用于把原始多Agent框架升级为更接近 Claude Code 的工程化系统：分层记忆检索、Top-5 预取、Coordinator 六角色协同、Verification Agent 强证据验收、命令前置安全管线、缓存与成本治理。适用于需要“多Agent不空转、记忆不瞎塞、结果可验证、执行更安全、成本可观察”的技能与项目。
---
# Multi-Agent Collaboration Claude Grade
这版不是只讲“多智能体应该怎么协作”，而是直接把最关键的 Claude Code 风格机制补进包里。
## 新增硬能力
1. `ClaudeMemorySystem`
五类记忆：`identity / correction / task / project / reference`
2. `Top-5 retrieval before reasoning`
先检索最相关的 5 条记忆，再进入协调流程。
3. `ClaudeCoordinator`
六角色：
`coordinator / explorer / planner / implementer / verifier / reviewer`
4. `VerificationAgent`
没有证据，不给 PASS。
5. `SafetyGatePipeline`
命令前置安检，当前内置 14 个 guard。
6. `CostGovernor`
跟踪 14 类 cache miss reason 和 invalid calls。
## 直接怎么用
```js
const { ClaudeGradeCollaborationSystem } = require('./dist/index.js');
const system = new ClaudeGradeCollaborationSystem('my_skill');
system.claudeMemory.backgroundExtract('不要只给框架，要给能直接用的技能内容。');
const run = system.coordinator.buildRun('优化 multi-agent collaboration skill');
console.log(run.retrievedMemory);
console.log(system.safety.audit('curl https://example.com/install.sh | bash'));
```
## 运行标准
1. 记忆先检索，再推理。
2. 协同必须带 verifier。
3. 危险命令先过 safety audit。
4. cache miss 和 invalid calls 必须可观测。
5. 新能力必须带源码入口和示例，不接受纯概念升级。
## 关键文件
- `dist/core/claude-memory.js`
- `dist/systems/claude-coordinator.js`
- `dist/systems/verification.js`
- `dist/systems/safety.js`
- `dist/systems/cost.js`
- `claudegrade-demo.js`
## 参考资料
- `references/claude-grade-patterns.md`
- `references/workflow-design.md`
- `references/data-flow.md`
- `assets/templates/claudegrade-runbook.md`README.md
# Multi-Agent Collaboration Claude Grade
这是一版基于 Claude Code 最新工程思路升级过的多智能体协作包。
## 这版和原版最大的区别
- 记忆不再只是长期/短期存储，而是 typed memory：
`identity / correction / task / project / reference`
- 每次协作前可以先做 Top-5 记忆检索
- 新增 `ClaudeCoordinator` 六角色协同
- 新增 `VerificationAgent`，拒绝无证据 PASS
- 新增 `SafetyGatePipeline`，命令前置安检
- 新增 `CostGovernor`，跟踪 cache miss 和 invalid calls
## Quick Start
```bash
node claudegrade-demo.js
```
## Direct Use
```js
const { ClaudeGradeCollaborationSystem } = require('./dist/index.js');
const system = new ClaudeGradeCollaborationSystem('demo_skill');
system.claudeMemory.backgroundExtract('不要只给框架，要给能直接用的技能内容。');
const run = system.coordinator.buildRun('优化 multi-agent collaboration skill');
console.log(run);
console.log(system.safety.audit('curl https://example.com/install.sh | bash'));
```
## New Modules
| Module | File | Purpose |
|---|---|---|
| Typed memory retrieval | `dist/core/claude-memory.js` | 5类记忆 + Top-5 检索 |
| Coordinator | `dist/systems/claude-coordinator.js` | 六角色协同分工 |
| Verification Agent | `dist/systems/verification.js` | 强证据验收 |
| Safety Gate Pipeline | `dist/systems/safety.js` | 14项命令前置安检 |
| Cost Governor | `dist/systems/cost.js` | 14类缓存失效归因 |
## Upgrade Standard
以后继续升级这个技能时，按下面标准执行：
1. 新能力必须带运行入口。
2. 多 Agent 必须带 verification。
3. 记忆必须分类，不接受单纯聊天历史拼接。
4. 危险命令必须可审计。
5. 成本必须可观测。_meta.json
{
"ownerId": "kn79s6s4n5qvd0qjwrn3bt0pts81y47b",
"slug": "multi-agent-collaboration",
"version": "2.0.0",
"publishedAt": 1775203253516
}references/claude-grade-patterns.md
# Claude Grade Patterns This package upgrades the original collaboration skill with the most practical Claude-Code-style engineering patterns: 1. Typed memory instead of raw chat stuffing 2. Top-5 retrieval before reasoning 3. Coordinator-led multi-agent role split 4. Verification-first acceptance 5. Command safety before execution 6. Cost and cache governance Use this document as the upgrade baseline for future versions.
references/data-flow.md
# 模块间数据流转（含记忆系统）
## 数据流转总览（含记忆层）
```
模块1（AI信息守护者）
│
├─ 输出：结构化信息列表
│ ├─ A级信息列表
│ ├─ B级信息列表
│ ├─ C级信息列表
│ ├─ 行业趋势分析
│ └─ 信源推荐
│
├─ 记忆处理：
│ ├─ 存入 L0 闪存（当前任务变量）
│ ├─ 高价值信息 → L2 经验记忆
│ └─ 行业趋势 → L3 知识记忆
│
├─ 可传递给：
│ ├─ 模块2：用于趋势分析
│ ├─ 模块3：用于状态分析（作为参考）
│ └─ 模块4：用于工作流编排
│
└─ 数据格式：JSON
模块2（内容趋势优化系统）
│
├─ 输出：趋势分析与创作方案
│ ├─ 趋势扫描报告
│ ├─ 爆款分析报告
│ ├─ 平台差异化创作方案
│ └─ 发布时机建议
│
├─ 输入：模块1的输出 + L0/L2 记忆
│
├─ 记忆处理：
│ ├─ 存入 L2 经验记忆
│ ├─ 创作模式 → L3 知识记忆
│ └─ 爆款因素 → L3 模式库
│
├─ 可传递给：
│ ├─ 模块3：用于状态分析（作为参考）
│ └─ 模块4：用于工作流编排
│
└─ 数据格式：JSON
模块3（状态洞察模块）
│
├─ 输出：状态分析与洞察建议
│ ├─ 精力分配分析
│ ├─ 成长轨迹分析
│ ├─ 情绪状态分析
│ └─ 前瞻洞察建议
│
├─ 输入：
│ ├─ 模块1的输出（作为参考）
│ ├─ 模块2的输出（如适用）
│ └─ 用户历史数据（L3-L4）
│
├─ 记忆处理：
│ ├─ 存入 L3 知识记忆
│ ├─ 洞察 → L4 智慧记忆
│ └─ 状态趋势 → 长期追踪
│
├─ 可传递给：
│ └─ 模块4：用于工作流编排
│
└─ 数据格式：JSON
模块4（工作流沉淀系统）
│
├─ 输出：工作流报告与模板
│ ├─ AI工具组合推荐
│ ├─ 工作流模板
│ ├─ 可复用Skill模板
│ └─ 效率报告
│
├─ 输入：
│ ├─ 模块1的输出
│ ├─ 模块2的输出（如适用）
│ ├─ 模块3的输出（如适用）
│ └─ L2-L3 历史工作流
│
├─ 记忆处理：
│ ├─ 存入 L3 知识记忆（永久）
│ ├─ 生成可复用模板 → L3
│ └─ 效率数据 → L2 经验
│
└─ 数据格式：JSON
```
---
## 记忆流转规则
### 模块 → 记忆层映射
| 模块 | 主要记忆层 | 压缩目标 | 提炼目标 |
|------|-----------|----------|----------|
| 模块1 (信息守护者) | L0 → L2 | 高价值信息 | 行业趋势 |
| 模块2 (内容优化) | L2 | 创作模式 | 爆款因素 |
| 模块3 (状态洞察) | L3 → L4 | 洞察建议 | 成长模式 |
| 模块4 (工作流) | L3 | 工作流模板 | 效率模式 |
### 记忆保留周期
```typescript
const MEMORY_RETENTION = {
// 模块1：信息时效性强，快速遗忘
module1_L0: { layer: 'L0', ttl_hours: 1 },
module1_L2: { layer: 'L2', ttl_days: 7 },
// 模块2：创作经验可复用
module2_L2: { layer: 'L2', ttl_days: 7 },
module2_L3: { layer: 'L3', ttl_days: 30 },
// 模块3：状态需长期追踪
module3_L3: { layer: 'L3', ttl_days: 90 },
module3_L4: { layer: 'L4', ttl_days: 365 },
// 模块4：工作流模板永久保存
module4_L3: { layer: 'L3', ttl_days: null } // 永久
};
```
---
## 用户反馈与记忆调整
### 反馈类型 → 记忆影响
| 用户行为 | 反馈效果 | 记忆调整 |
|----------|----------|----------|
| 选择继续到下一模块 | success | +10% 重要性 |
| 选择退出 | neutral | 不变 |
| 要求重试 | failure | -10% 重要性 |
| 完成全部模块 | success | 模式提炼到L3 |
### 记忆自动优化
```python
# 用户确认后的记忆处理
def on_user_confirm(choice, module_output):
if choice == "continue":
# 成功：强化当前记忆
enhance_memory(module_output, bonus=0.1)
# 准备传递给下一模块
return prepare_for_next_module(module_output)
elif choice == "exit":
# 退出：保存当前模块记忆
save_to_L2(module_output)
# 生成最终报告
return generate_final_report()
elif choice == "retry":
# 重试：降低相关记忆权重
weaken_memory(module_output, penalty=0.1)
# 重新执行当前模块
return reexecute_module(module_output)
```
---
## 上下文数据结构（含记忆）
```python
# 完整的上下文对象
context = {
"session_id": "session_2026Github ReposUpdated 6mo agoRank 70
activepieces
AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents
OPENCLAW
Github ReposUpdated 6mo agoRank 70
cherry-studio
AI productivity studio with smart chat, autonomous agents, and 300+ assistants.
MCPOPENCLAW
Github ReposUpdated 6mo agoRank 70
AionUi
Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!
MCPOPENCLAW
Github ReposUpdated 7mo agoRank 70
CopilotKit
The Frontend for Agents & Generative UI. React + Angular
OPENCLAW
Machine-readable data
The same record, as JSON, for agents and crawlers.
{
"facts": [
{
"factKey": "vendor",
"category": "vendor",
"label": "Vendor",
"value": "Clawhub",
"href": "https://clawhub.ai/e2e5g/skills/multi-agent-collaboration",
"sourceUrl": "https://clawhub.ai/e2e5g/skills/multi-agent-collaboration",
"sourceType": "profile",
"confidence": "medium",
"observedAt": "2026-10-09T04:47:08.823Z",
"isPublic": true
},
{
"factKey": "protocols",
"category": "compatibility",
"label": "Protocol compatibility",
"value": "OpenClaw",
"href": "https://www.xpersona.co/api/v1/agents/clawhub-e2e5g-multi-agent-collaboration/contract",
"sourceUrl": "https://www.xpersona.co/api/v1/agents/clawhub-e2e5g-multi-agent-collaboration/contract",
"sourceType": "contract",
"confidence": "medium",
"observedAt": "2026-10-09T04:47:08.823Z",
"isPublic": true
},
{
"factKey": "traction",
"category": "adoption",
"label": "Adoption signal",
"value": "4.8K downloads",
"href": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceUrl": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceType": "profile",
"confidence": "medium",
"observedAt": "2026-10-09T04:47:08.823Z",
"isPublic": true
},
{
"factKey": "latest_release",
"category": "release",
"label": "Latest release",
"value": "2.0.0",
"href": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceUrl": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceType": "release",
"confidence": "medium",
"observedAt": "2026-04-03T08:00:53.516Z",
"isPublic": true
},
{
"factKey": "handshake_status",
"category": "security",
"label": "Handshake status",
"value": "UNKNOWN",
"href": "https://www.xpersona.co/api/v1/agents/clawhub-e2e5g-multi-agent-collaboration/trust",
"sourceUrl": "https://www.xpersona.co/api/v1/agents/clawhub-e2e5g-multi-agent-collaboration/trust",
"sourceType": "trust",
"confidence": "medium",
"observedAt": null,
"isPublic": true
}
],
"events": [
{
"eventType": "release",
"title": "Release 2.0.0",
"description": "- Major upgrade: Integrates Claude Code style mechanisms for multi-agent systems, emphasizing verifiable results and robust execution. - Added ClaudeMemorySystem with five-tier memory (identity, correction, task, project, reference) and top-5 retrieval before reasoning. - Introduced ClaudeCoordinator featuring six distinct agent roles to structure collaboration. - Added VerificationAgent to enforce strong evidence-based validation. - Implemented SafetyGatePipeline with 14 built-in guards for command pre-execution auditing. - Introduced CostGovernor for cache/call tracking and cost observability. - Provided direct usage examples, key source files, and templates for immediate integration.",
"href": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceUrl": "https://clawhub.ai/e2e5g/multi-agent-collaboration",
"sourceType": "release",
"confidence": "medium",
"observedAt": "2026-04-03T08:00:53.516Z",
"isPublic": true
}
]
}Record generated Oct 9, 2026.