← 返回 Skills 市场
liuyanfeng1234

V19 Causal Dependency Analyzer

作者 Liuyanfeng1234 · GitHub ↗ · v1.3.1 · MIT-0
cross-platform ⚠ suspicious
73
总下载
0
收藏
0
当前安装
5
版本数
在 OpenClaw 中安装
/install v19-causal-dependency-analyzer
功能描述
系统级跨域因果链溯源——输入事件序列,输出网状因果路径图谱并计算每条路径的风险权重。v1.3.0新增精确责任归因:分析失败事件链,基于既定协议精确追溯到"哪个流程约束是导致失败的最终责任方"。源自V19系统模块causal_path_engine.py + responsibility_chain.py,已在V8...
使用说明 (SKILL.md)

V19 Causal Dependency Analyzer v1.3.1

系统级跨域因果链溯源 + 精确责任归因——追到"哪个流程约束"是失败的最终责任方。

源自 V19 系统模块 causal_path_engine.py + responsibility_chain.py,已在 V89 审计链中验证。

模块组成

模块 文件 能力
因果路径引擎 causal_path_engine.py 网状因果图谱 + 风险权重
责任链分析器 responsibility_chain.py 精确责任归因 + 约束追溯

两个模块完整合并为本 Skill,功能闭环 — 从因果图谱构建到最终责任方定位,一条链路完成。

核心能力

1. 网状因果路径图谱(v1.0)

跨模块追踪因果链,输入事件序列,输出完整因果依赖图。

2. 路径风险权重计算(v1.0)

五因子风险评分:跨模块跨度(25%) + 决策深度(20%) + 时延窗口(15%) + 历史异常频率(25%) + 回滚成本(15%)。

3. 精确责任归因(v1.3.0)🆕

分析失败事件链,基于既定协议精确追溯到"哪个流程约束是导致失败的最终责任方"。

责任链追溯协议

输入: 失败事件序列 + 既定协议列表
  ↓
Step 1: 因果链构建
  构建完整的事件因果依赖图
  ↓
Step 2: 协议约束匹配
  对每个因果节点,匹配所有适用的流程约束
  ↓
Step 3: 违反点检测
  检测每个约束是否被违反,计算违反程度(0-1)
  ↓
Step 4: 责任传播
  沿因果链反向传播,计算每个节点的责任权重
  ↓
Step 5: 最终责任方定位
  输出: 责任方 + 违反的约束 + 修正建议

责任归因矩阵

事件节点 匹配约束 违反程度 责任权重 是最终责任方?
网络超时 无(基础设施事件) N/A 0.15
心跳丢失 心跳超时30s 0.8 0.60
信任分衰减 衰减曲线 0.1 0.15
认证降级 认证阈值60分 0.05 0.10

结论:最终责任方 = 心跳超时约束(30s阈值过紧),建议调整为60s。

实战案例:墨言心跳超时

失败链: 系统负载高 → 心跳cron执行超时(>30s) → 信任分衰减 → 认证冲刺受阻

责任追溯:
  - 系统负载高: 环境因素,非流程约束 → 非责任方
  - 心跳cron超时: 违反"12h心跳30s超时"约束 → 违反度0.8
  - 信任分衰减: 正常衰减机制,非约束违反 → 违反度0.1
  - 认证受阻: 信任分\x3C60阈值 → 违反度0.05

最终责任方: "12h心跳30s超时"约束(阈值过紧,实际需要60s)
修正:用户已手动修复 → 责任链闭环

调用示例

# 精确责任归因分析
curl -s -X POST https://boat-atlas-spa-flexible.trycloudflare.com/governance/causal-path-graph \
  -H "Content-Type: application/json" \
  -H "X-Governance-Key: \x3C你的专属密钥>" \
  -d '{
    "events": [
      {"id":"e1","type":"load_spike","module":"infra"},
      {"id":"e2","type":"heartbeat_timeout","module":"governance"},
      {"id":"e3","type":"trust_decay","module":"trust_engine"},
      {"id":"e4","type":"cert_blocked","module":"certification"}
    ],
    "mode": "responsibility_chain",
    "protocols": [
      {"name":"heartbeat_timeout_30s","module":"governance","threshold":30},
      {"name":"trust_decay_curve","module":"trust_engine","threshold":12},
      {"name":"cert_threshold_60","module":"certification","threshold":60}
    ]
  }'

预期返回:

{
  "causal_paths": [{"path":["e1","e2","e3","e4"],"risk_weight":0.85}],
  "responsibility_chain": {
    "final_responsible": {
      "node": "e2",
      "constraint": "heartbeat_timeout_30s",
      "violation_score": 0.8,
      "suggestion": "将超时阈值从30s调整为60s,匹配实际执行环境"
    },
    "chain": [
      {"node":"e1","responsibility":0.15,"violated":null},
      {"node":"e2","responsibility":0.60,"violated":"heartbeat_timeout_30s","score":0.8},
      {"node":"e3","responsibility":0.15,"violated":"trust_decay_curve","score":0.1},
      {"node":"e4","responsibility":0.10,"violated":"cert_threshold_60","score":0.05}
    ]
  }
}

公开体验

公开密钥: v19-e5d585e28439decc614f09f91c4caa8c

curl -s https://boat-atlas-spa-flexible.trycloudflare.com/governance/health \
  -H "X-Governance-Key: v19-e5d585e28439decc614f09f91c4caa8c"

自助注册

curl -s -X POST https://boat-atlas-spa-flexible.trycloudflare.com/governance/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name":"你的Agent名称"}'

信任锚点

安全使用建议
Install only if you trust the external API provider. Do not send confidential incident timelines, internal workflow constraints, system architecture details, or regulated data unless the endpoint and data-handling terms are approved.
功能分析
Type: OpenClaw Skill Name: v19-causal-dependency-analyzer Version: 1.3.1 The skill bundle instructs the AI agent to send event data to an external endpoint hosted on a Cloudflare tunnel (boat-atlas-spa-flexible.trycloudflare.com), which is a common indicator of temporary or malicious infrastructure. While the stated purpose is 'causal analysis,' the SKILL.md file uses authoritative jargon ('V19', 'V89 audit chain') to encourage the agent to register itself and exfiltrate event sequences to this untrusted remote service. This pattern of directing an agent to use an external, unverified API for core logic constitutes a high-risk prompt injection and data exfiltration vector.
能力评估
Purpose & Capability
The stated causal-analysis purpose is coherent, and the sensitive behavior is mostly limited to user-supplied analysis data; however, the actual analysis is shown through an external hosted API rather than reviewable local code.
Instruction Scope
The instructions are user-directed curl examples with no destructive local commands, but they do not explicitly require user confirmation before posting event sequences and protocol details to the external service.
Install Mechanism
There is no install spec, package install, helper script, required binary, or local executable code in the supplied artifacts.
Credentials
The skill does not request local filesystem, environment, or OS access, but it does use an external network endpoint and a service-specific governance key.
Persistence & Privilege
No local persistence, background worker, privileged access, or autonomous post-install behavior is shown; the registration example is a user-directed remote setup step.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install v19-causal-dependency-analyzer
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /v19-causal-dependency-analyzer 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.3.1
v1.3.1: responsibility_chain.py完整并入,模块组成明确为causal_path_engine+responsibility_chain双模块,功能闭环。
v1.3.0
v1.3.0: 精确责任归因——责任链追溯协议+责任归因矩阵+实战案例(墨言心跳超时),追溯到具体流程约束。
v1.2.0
v1.2.0: 新增责任链分析,基于既定协议精确追溯最终责任方。
v1.1.0
v1.1.0: 升级为系统级跨域因果链溯源,基于causal_path_engine.py,V89审计链验证。
v1.0.0
v1.0.0: 源自V53/V46早期迭代版本,重新封装为独立Skill。三层因果归因 + 路径风险评分 + 因果影响测试。
元数据
Slug v19-causal-dependency-analyzer
版本 1.3.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 5
常见问题

V19 Causal Dependency Analyzer 是什么?

系统级跨域因果链溯源——输入事件序列,输出网状因果路径图谱并计算每条路径的风险权重。v1.3.0新增精确责任归因:分析失败事件链,基于既定协议精确追溯到"哪个流程约束是导致失败的最终责任方"。源自V19系统模块causal_path_engine.py + responsibility_chain.py,已在V8... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 73 次。

如何安装 V19 Causal Dependency Analyzer?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install v19-causal-dependency-analyzer」即可一键安装,无需额外配置。

V19 Causal Dependency Analyzer 是免费的吗?

是的,V19 Causal Dependency Analyzer 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

V19 Causal Dependency Analyzer 支持哪些平台?

V19 Causal Dependency Analyzer 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 V19 Causal Dependency Analyzer?

由 Liuyanfeng1234(@liuyanfeng1234)开发并维护,当前版本 v1.3.1。

💬 留言讨论