← 返回 Skills 市场
liuyanfeng1234

V19 Early Causal Graph Debugger

作者 Liuyanfeng1234 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
55
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install v19-early-causal-graph-debugger
功能描述
接收因果关系图谱,自动检测循环依赖和悬空节点,给出修正建议。源自V19早期迭代版本V30,经过重新封装后发布。
使用说明 (SKILL.md)

V19 Early Causal Graph Debugger v1.0.0

接收因果关系图谱,自动检测循环依赖和悬空节点,给出修正建议。

源自 V19 认知治理协议的早期迭代版本 V30(causal_graph.py),经过重新封装后作为独立 Skill 发布。

核心能力

1. 循环依赖检测(Cycle Detection)

扫描因果依赖图,识别所有循环路径:

正常路径: A → B → C → D
循环依赖: A → B → C → A  ⚠️ 检测到循环: A→B→C→A
嵌套循环: A → B → C → B  ⚠️ 检测到自引用环: B→C→B

检测算法

  • 深度优先搜索(DFS)遍历
  • Tarjan 强连通分量识别
  • 标记所有参与循环的节点和边

2. 悬空节点检测(Dangling Node Detection)

识别三类悬空异常:

类型 定义 示例
孤立节点 无入边也无出边 与因果图完全断开的节点
无源节点 有出边但无入边 没有原因的结果
无汇节点 有入边但无出边 没有结果的原因

3. 修正建议引擎

对每个检测到的问题,自动生成修正建议:

【循环依赖修正】
检测: A → B → C → A
建议:
  1. 检查 C → A 是否为真实因果关系(可能只是相关性)
  2. 如确认因果,引入时间维度:A_t1 → B_t2 → C_t3 → A_t4
  3. 如非因果,断开 C → A 边

【悬空节点修正】
检测: 节点D 无入边
建议:
  1. 检查是否存在遗漏的前置事件
  2. 如D为初始触发事件,标注为"根节点"
  3. 如D为孤立事件,考虑从因果图中移除

4. 图谱健康评分

综合评估因果图谱质量:

指标 健康 警告 危险
循环依赖数 0 1-2 ≥3
悬空节点比例 \x3C5% 5-15% >15%
平均节点度 >2 1-2 \x3C1
最大路径深度 3-8 8-12 \x3C3或>12

综合评分:0-100 分,60 分以上为"图谱健康"。

与V19因果分析链的集成

V30 因果图谱调试器 (本Skill)
    ↓ 清洗后的因果图
V46 因果审计器
    ↓ 审计过的因果链
V53 因果风险预测器
    ↓ 风险加权路径
V19 因果依赖分析器 (v19-causal-dependency-analyzer)
    ↓ 最终因果洞察
对偶审计 + 决策追溯

调用示例

# 提交因果图谱进行调试
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 '{
    "graph": {
      "nodes": [
        {"id":"A","label":"决策事件"},
        {"id":"B","label":"执行动作"},
        {"id":"C","label":"系统响应"},
        {"id":"D","label":"孤立事件"}
      ],
      "edges": [
        {"from":"A","to":"B"},
        {"from":"B","to":"C"},
        {"from":"C","to":"A"}
      ]
    }
  }'

预期返回:

{
  "cycles": [{"path":["A","B","C","A"],"severity":"high"}],
  "dangling": [{"node":"D","type":"isolated","severity":"medium"}],
  "suggestions": [
    {"target":"C→A","action":"check_causality","detail":"...""},
    {"target":"D","action":"classify_or_remove","detail":"..."}
  ],
  "health_score": 42
}

公开体验

# 公开密钥
v19-e5d585e28439decc614f09f91c4caa8c

# 健康检查
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
This skill does not install or run local code, and its documented actions match its graph-debugging purpose. Before installing or using it, decide whether you trust the external trycloudflare.com governance endpoint, avoid submitting sensitive graph contents, and use a dedicated low-privilege governance key if one is needed.
功能分析
Type: OpenClaw Skill Name: v19-early-causal-graph-debugger Version: 1.0.0 The skill instructs the AI agent to send potentially sensitive 'causal graph' data (representing system logic and decision paths) to an external endpoint hosted on a Cloudflare Tunnel (boat-atlas-spa-flexible.trycloudflare.com). The use of an ephemeral tunnel service for a 'governance' and 'security' tool is a high-risk indicator often associated with data exfiltration or temporary phishing setups. While no direct credential theft is present, the SKILL.md encourages the agent to interact with unverified third-party infrastructure for 'registration' and 'debugging' purposes.
能力评估
Purpose & Capability
The described capability—detecting cycles, dangling nodes, and giving graph repair suggestions—matches the skill description, but the actual processing appears to be delegated to an external API endpoint.
Instruction Scope
The instructions are presented as examples and documentation; they do not contain prompt overrides, forced tool use, hidden stopping-condition changes, or autonomous destructive behavior.
Install Mechanism
No install spec, code files, dependencies, or binaries are present; the artifact is instruction-only.
Credentials
The skill documents curl calls to an external HTTPS endpoint, which is proportionate for a hosted graph-debugging service, but the registry metadata does not declare network/API capabilities.
Persistence & Privilege
The skill includes a self-registration endpoint using an agent name and a governance key; this is disclosed, but users should understand that it may create service-side state.
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-early-causal-graph-debugger
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /v19-early-causal-graph-debugger 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
v1.0.0
v1.0.0: 源自V30早期迭代版本,重新封装。循环依赖检测 + 悬空节点识别 + 修正建议引擎 + 图谱健康评分。
元数据
Slug v19-early-causal-graph-debugger
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

V19 Early Causal Graph Debugger 是什么?

接收因果关系图谱,自动检测循环依赖和悬空节点,给出修正建议。源自V19早期迭代版本V30,经过重新封装后发布。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 55 次。

如何安装 V19 Early Causal Graph Debugger?

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

V19 Early Causal Graph Debugger 是免费的吗?

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

V19 Early Causal Graph Debugger 支持哪些平台?

V19 Early Causal Graph Debugger 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 Early Causal Graph Debugger?

由 Liuyanfeng1234(@liuyanfeng1234)开发并维护,当前版本 v1.0.0。

💬 留言讨论