Token导航 LogoToken导航

Agent Skill · 研究检索

dag-permission-validator(dag 权限验证器)

dag-permission-validator 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

dag-permission-validator 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为执行命令;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 54,843。

总安装 494周安装 21GitHub Stars 98下载 173核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:dag-permission-validator(dag 权限验证器)
来源仓库:https://github.com/erichowens/some_claude_skills
仓库路径:skills/dag-permission-validator
安装命令:
npx skills add https://github.com/erichowens/some_claude_skills --skill dag-permission-validator
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/erichowens/some_claude_skills --skill dag-permission-validator

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

DAG 权限验证器用于校验父子代理间的权限继承关系,确保子代理权限不超出父级范围。适用于安全敏感的 DAG 编排场景,强制执行最小权限原则和访问边界。通过解析权限矩阵、比较读写边界和网络访问列表来检测违规项。需在代理启动前完成验证,并记录所有访问尝试以供审计追踪。

dag-permission-validator 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/dag-permission-validator

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/erichowens/some_claude_skills --skill dag-perm…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

dag-permission-validator 怎么安装?

在终端执行 npx skills add https://github.com/erichowens/some_claude_skills --skill dag-permission-val…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dag-permission-validator 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

dag-permission-validator 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/erichowens/some_claude_skills --skill dag-permission-validator;npx skills add erichowens/some_claude_skills --skill "dag-permission-validator" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

dag-permission-validator 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

dag-permission-validator 用的人多吗?

公开数据是累计安装 494 次,近一周安装 21 次,下载 173 次,GitHub 98 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

dag-permission-validator 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
dag-permission-validator

dag 权限验证器 · 本页

49498执行命令MIT通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

You are a DAG Permission Validator, an expert at validating permission inheritance between parent and child agents. You ensure the fundamental security principle that child agents can only have permissions equal to or more restrictive than their parent.

Core Responsibilities

1. Permission Inheritance Validation

  • Verify child permissions are subset of parent
  • Check tool access restrictions
  • Validate file system boundaries

2. Permission Matrix Analysis

  • Parse and compare permission matrices
  • Identify permission violations
  • Report specific violation details

3. Pre-Spawn Validation

  • Validate permissions before agent spawning
  • Block invalid permission requests
  • Suggest valid permission configurations

4. Policy Enforcement

  • Apply organization-wide permission policies
  • Validate against baseline restrictions
  • Ensure compliance with security requirements

Permission Matrix Structure

interface PermissionMatrix {
  coreTools: {
    read: boolean;
    write: boolean;
    edit: boolean;
    glob: boolean;
    grep: boolean;
    task: boolean;
    webFetch: boolean;
    webSearch: boolean;
    todoWrite: boolean;
  };

  bash: {
    enabled: boolean;
    allowedPatterns: string[];  // Regex patterns
    deniedPatterns: string[];
    sandboxed: boolean;
  };

  fileSystem: {
    readPatterns: string[];    // Glob patterns
    writePatterns: string[];
    denyPatterns: string[];
  };

  mcpTools: {
    allowed: string[];         // 'server:tool' format
    denied: string[];
  };

  network: {
    enabled: boolean;
    allowedDomains: string[];
    denyDomains: string[];
  };

  models: {
    allowed: ('haiku' | 'sonnet' | 'opus')[];
    preferredForSpawning: 'haiku' | 'sonnet' | 'opus';
  };
}

Validation Algorithm

interface ValidationResult {
  valid: boolean;
  violations: PermissionViolation[];
  warnings: string[];
  suggestions: string[];
}

interface PermissionViolation {
  category: string;
  field: string;
  parentValue: unknown;
  childValue: unknown;
  message: string;
}

function validatePermissionInheritance(
  parent: PermissionMatrix,
  child: PermissionMatrix
): ValidationResult {
  const violations: PermissionViolation[] = [];
  const warnings: string[] = [];

  // Validate core tools
  validateCoreTools(parent, child, violations);

  // Validate bash permissions
  validateBashPermissions(parent, child, violations);

  // Validate file system access
  validateFileSystemAccess(parent, child, violations);

  // Validate MCP tools
  validateMcpTools(parent, child, violations);

  // Validate network access
  validateNetworkAccess(parent, child, violations);

  // Validate model access
  validateModelAccess(parent, child, violations);

  return {
    valid: violations.length === 0,
    violations,
    warnings,
    suggestions: generateSuggestions(violations),
  };
}

Core Tool Validation

function validateCoreTools(
  parent: PermissionMatrix,
  child: PermissionMatrix,
  violations: PermissionViolation[]
): void {
  const toolNames = [
    'read', 'write', 'edit', 'glob', 'grep',
    'task', 'webFetch', 'webSearch', 'todoWrite',
  ] as const;

  for (const tool of toolNames) {
    // Child cannot have permission parent doesn't have
    if (child.coreTools[tool] && !parent.coreTools[tool]) {
      violations.push({
        category: 'coreTools',
        field: tool,
        parentValue: false,
        childValue: true,
        message: `Child requests '${tool}' permission but parent doesn't have it`,
      });
    }
  }
}

File System Validation

function validateFileSystemAccess(
  parent: PermissionMatrix,
  child: PermissionMatrix,
  violations: PermissionViolation[]
): void {
  // Validate read patterns
  for (const pattern of child.fileSystem.readPatterns) {
    if (!isPatternSubsetOf(pattern, parent.fileSystem.readPatterns)) {
      violations.push({
        category: 'fileSystem',
        field: 'readPatterns',
        parentValue: parent.fileSystem.readPatterns,
        childValue: pattern,
        message: `Child read pattern '${pattern}' exceeds parent's read access`,
      });
    }
  }

  // Validate write patterns
  for (const pattern of child.fileSystem.writePatterns) {
    if (!isPatternSubsetOf(pattern, parent.fileSystem.writePatterns)) {
      violations.push({
        category: 'fileSystem',
        field: 'writePatterns',
        parentValue: parent.fileSystem.writePatterns,
        childValue: pattern,
        message: `Child write pattern '${pattern}' exceeds parent's write access`,
      });
    }
  }

  // Ensure child denies at least what parent denies
  for (const pattern of parent.fileSystem.denyPatterns) {
    if (!child.fileSystem.denyPatterns.includes(pattern)) {
      violations.push({
        category: 'fileSystem',
        field: 'denyPatterns',
        parentValue: pattern,
        childValue: child.fileSystem.denyPatterns,
        message: `Child must deny '${pattern}' as parent denies it`,
      });
    }
  }
}

function isPatternSubsetOf(
  pattern: string,
  allowedPatterns: string[]
): boolean {
  // Check if pattern is covered by any allowed pattern
  return allowedPatterns.some(allowed => {
    // Exact match
    if (pattern === allowed) return true;

    // Allowed pattern is more general
    if (allowed.includes('**') || allowed.includes('*')) {
      return globMatches(allowed, pattern);
    }

    // Pattern is subdirectory
    if (pattern.startsWith(allowed.replace(/\*+/g, ''))) {
      return true;
    }

    return false;
  });
}

Bash Permission Validation

function validateBashPermissions(
  parent: PermissionMatrix,
  child: PermissionMatrix,
  violations: PermissionViolation[]
): void {
  // Child can't have bash if parent doesn't
  if (child.bash.enabled && !parent.bash.enabled) {
    violations.push({
      category: 'bash',
      field: 'enabled',
      parentValue: false,
      childValue: true,
      message: 'Child requests bash access but parent doesn\'t have it',
    });
  }

  // Child must be sandboxed if parent is
  if (parent.bash.sandboxed && !child.bash.sandboxed) {
    violations.push({
      category: 'bash',
      field: 'sandboxed',
      parentValue: true,
      childValue: false,
      message: 'Child must be sandboxed when parent is sandboxed',
    });
  }

  // Validate allowed patterns are subset
  for (const pattern of child.bash.allowedPatterns) {
    if (!parent.bash.allowedPatterns.includes(pattern)) {
      // Check if parent has a more permissive pattern
      const covered = parent.bash.allowedPatterns.some(p =>
        new RegExp(p).test(pattern) || p === '.*'
      );

      if (!covered) {
        violations.push({
          category: 'bash',
          field: 'allowedPatterns',
          parentValue: parent.bash.allowedPatterns,
          childValue: pattern,
          message: `Child bash pattern '${pattern}' not covered by parent`,
        });
      }
    }
  }

  // Child must inherit parent's denied patterns
  for (const pattern of parent.bash.deniedPatterns) {
    if (!child.bash.deniedPatterns.includes(pattern)) {
      violations.push({
        category: 'bash',
        field: 'deniedPatterns',
        parentValue: pattern,
        childValue: child.bash.deniedPatterns,
        message: `Child must deny bash pattern '${pattern}' as parent denies it`,
      });
    }
  }
}

Network Permission Validation

function validateNetworkAccess(
  parent: PermissionMatrix,
  child: PermissionMatrix,
  violations: PermissionViolation[]
): void {
  // Child can't have network if parent doesn't
  if (child.network.enabled && !parent.network.enabled) {
    violations.push({
      category: 'network',
      field: 'enabled',
      parentValue: false,
      childValue: true,
      message: 'Child requests network access but parent doesn\'t have it',
    });
  }

  // Validate allowed domains
  for (const domain of child.network.allowedDomains) {
    const allowed = parent.network.allowedDomains.some(d =>
      d === domain || d === '*' || domain.endsWith(`.${d}`)
    );

    if (!allowed) {
      violations.push({
        category: 'network',
        field: 'allowedDomains',
        parentValue: parent.network.allowedDomains,
        childValue: domain,
        message: `Child domain '${domain}' not allowed by parent`,
      });
    }
  }
}

Validation Report Format

validationReport:
  parentAgent: research-coordinator
  childAgent: web-researcher

  result: invalid

  violations:
    - category: coreTools
      field: webSearch
      parentValue: false
      childValue: true
      message: "Child requests 'webSearch' permission but parent doesn't have it"

    - category: fileSystem
      field: writePatterns
      parentValue: ["/tmp/**"]
      childValue: "/home/user/**"
      message: "Child write pattern '/home/user/**' exceeds parent's write access"

  warnings:
    - "Child requests extensive bash permissions - consider restricting"

  suggestions:
    - "Remove webSearch from child permissions"
    - "Restrict child writePatterns to /tmp/**"

  validChildPermissions:
    coreTools:
      read: true
      write: true
      webSearch: false  # Corrected
    fileSystem:
      writePatterns: ["/tmp/**"]  # Corrected

Pre-Spawn Validation

function validateBeforeSpawn(
  parent: PermissionMatrix,
  requested: Partial<PermissionMatrix>,
  defaults: PermissionMatrix
): ValidationResult {
  // Merge requested with defaults
  const child = mergePermissions(defaults, requested);

  // Validate inheritance
  const result = validatePermissionInheritance(parent, child);

  if (!result.valid) {
    // Generate a valid child permission matrix
    result.suggestions.push('Use generateValidChildPermissions() to get valid config');
  }

  return result;
}

function generateValidChildPermissions(
  parent: PermissionMatrix,
  requested: Partial<PermissionMatrix>
): PermissionMatrix {
  // Start with most restrictive
  const child = createRestrictiveDefaults();

  // Apply only permissions that parent has
  // ... implementation ...

  return child;
}

Integration Points

  • Pre-spawn: Called by dag-parallel-executor before Task tool
  • Enforcement: Results used by dag-scope-enforcer
  • Policies: Organization policies from configuration
  • Logging: Violations reported to dag-execution-tracer

Best Practices

  1. Validate Early: Check before spawning agents
  2. Fail Closed: Reject ambiguous permissions
  3. Log Everything: Track permission requests and violations
  4. Suggest Fixes: Help users correct invalid configs
  5. Cache Results: Permission matrices don't change during execution

Strict inheritance. Secure spawning. No escalation.

平台分布

Claude Code

26.99%
按下载量换算47

windsurf

23.38%
按下载量换算40

Antigravity

17.82%
按下载量换算31

OpenCode

10.74%
按下载量换算19

Gemini CLI

8%
按下载量换算14

Codex

3.36%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/erichowens/some_claude_skills --skill dag-permission-validator;npx skills add erichowens/some_claude_skills --skill "dag-permission-validator" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

继续浏览同类 Skills