Token导航 LogoToken导航

Agent Skill · 研究检索

ark-vulnerability-fixer(方舟漏洞修复器)

用于辅助安全审计、权限检查、凭据风险、认证流程和常见漏洞排查。它适合让 Agent 梳理敏感配置、检查依赖风险、分析鉴权逻辑或生成安全复核清单。使用时不能把工具输出直接当最终结论,涉及密钥、令牌、用户数据或生产系统时,应先确认最小权限、脱敏方式和操作边界。

ark-vulnerability-fixer 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,权限为需要联网;许可证为 MIT,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 44,563。

总安装 906周安装 37GitHub Stars 373下载 290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:ark-vulnerability-fixer(方舟漏洞修复器)
来源仓库:https://github.com/mckinsey/agents-at-scale-ark
仓库路径:skills/ark-vulnerability-fixer
安装命令:
npx skills add https://github.com/mckinsey/agents-at-scale-ark --skill ark-vulnerability-fixer
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/mckinsey/agents-at-scale-ark --skill ark-vulnerability-fixer

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。

ark-vulnerability-fixer 针对特定 CVE 提供补丁研究、影响分析与修复工作流指导。适合在已知漏洞编号(如 CVE-2025-55183)时快速定位 Ark 代码中的风险点并提供修复建议。输出包含 PoC 验证方法、依赖升级路径与回归测试要点,支持自动化 PR 生成。

使用前应交叉验证 CVE 数据库信息,避免依赖单一来源导致误判或遗漏上下文。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/ark-vulnerability-fixer

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/mckinsey/agents-at-scale-ark --skill ark-vulne…;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

    它需要需要联网,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

ark-vulnerability-fixer 怎么安装?

在终端执行 npx skills add https://github.com/mckinsey/agents-at-scale-ark --skill ark-vulnerability-f…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ark-vulnerability-fixer 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。

ark-vulnerability-fixer 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

ark-vulnerability-fixer 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。

ark-vulnerability-fixer 用的人多吗?

公开数据是累计安装 906 次,近一周安装 37 次,下载 290 次,GitHub 373 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。

ark-vulnerability-fixer 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

先想清楚的地方

  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。

不太适合

  • 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
ark-vulnerability-fixer

方舟漏洞修复器 · 本页

906373需要联网MIT有提醒

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

Ark Vulnerability Fixer

Provides CVE-specific research tools and security patch workflows for fixing vulnerabilities in Ark.

When to use this skill

Use this skill when:

  • User mentions a specific CVE number (e.g., "Fix CVE-2025-55183 in Ark")
  • User reports a security vulnerability that needs patching
  • You need CVE database information
  • You need security-focused PR templates

Note: This skill is typically used by the ark-security-patcher agent as part of a complete workflow:

  1. CVE research (this skill + research skill)
  2. Codebase analysis (this skill + analysis skill)
  3. Mitigation planning (this skill)
  4. Repository cloning and fix implementation
  5. Testing (optionally with setup skill)
  6. PR creation (this skill)

This skill complements the research, analysis, and setup skills for a complete end-to-end vulnerability fixing workflow.

CVE Research

CVE API Integration

Fetch official CVE data from the CIRCL CVE database:

# Fetch CVE details
curl -s "https://cve.circl.lu/api/cve/CVE-2025-55183" | python3 -m json.tool

The API provides:

  • Official CVE description
  • CVSS scores and severity ratings
  • References to security advisories
  • Affected products and version ranges
  • CWE categorization
  • Available patches and fixes

CVE Research Checklist

For each CVE, gather:

  • Official CVE description and CVSS score
  • Vendor security advisory
  • GitHub security advisory (if applicable)
  • Patch or fix documentation
  • Affected version range
  • Recommended version or workaround

Tip: Use the research skill for web searches to find vendor advisories and GitHub security alerts.

Dependency Analysis

Identifying Vulnerable Dependencies

Once you have CVE details, search Ark's dependencies:

cd /tmp/ark-analysis  # Use analysis skill to clone first

# Go dependencies
grep "package-name" go.mod go.sum
go list -m all | grep "package-name"

# Node.js dependencies
find . -name "package.json" -exec grep -l "package-name" {} \;
npm list package-name  # If in a node project

# Python dependencies
find . -name "requirements.txt" -o -name "pyproject.toml" | xargs grep "package-name"

# Docker base images
find . -name "Dockerfile" | xargs grep "FROM"

Assessing Impact

Consider Ark's specific context:

  • Deployment model: Kubernetes operator in cluster
  • Network exposure: Services typically internal to cluster
  • Trust boundary: Often in trusted environments
  • Attack vectors: What's realistic given Ark's architecture?

Tip: Use the analysis skill to understand Ark's architecture and service boundaries.

Mitigation Strategy

Presenting Options to User

CRITICAL: Always present mitigation options and wait for user approval before making changes.

Use this template to present findings:

## Security Vulnerability Analysis

### Vulnerability Details
- **CVE**: CVE-YYYY-NNNNN (or "Generic: [description]")
- **Severity**: [Critical/High/Medium/Low] (CVSS: [score])
- **Component**: [Library/package/framework]
- **Description**: [Clear explanation]

### Impact on Ark
- **Affected Services**: [List services/components]
- **Current Version**: [Version in use]
- **Vulnerable Versions**: [Range]
- **Attack Vector**: [How exploitable]
- **Risk Assessment**: [Realistic risk for Ark deployments]

### Mitigation Options

#### Option 1: [Recommended approach] (RECOMMENDED)
- **Action**: Update [component] from v[X] to v[Y]
- **Changes Required**: [Files to modify]
- **Testing Strategy**: [How to verify]
- **Impact**: [Breaking changes, if any]
- **Pros**: [Benefits]
- **Cons**: [Downsides]

#### Option 2: [Alternative approach]
- **Action**: [Alternative fix]
- **Changes Required**: [What changes]
- **Testing Strategy**: [How to verify]
- **Impact**: [Breaking changes, if any]
- **Pros**: [Benefits]
- **Cons**: [Downsides]

### Recommendation
Based on [evidence sources], I recommend **Option 1** because:
1. [Primary reason]
2. [Secondary reason]

### Next Steps
Would you like to proceed with this mitigation?

### Sources
- [CVE Database](https://cve.circl.lu/cve/CVE-YYYY-NNNNN)
- [Vendor Advisory](URL)

STOP AND WAIT for user approval before implementing.

Repository Setup for Fixes

Cloning for Development

After user approves the mitigation, clone Ark for making changes:

# Clone the repository
git clone git@github.com:mckinsey/agents-at-scale-ark.git
cd agents-at-scale-ark

# Create a security fix branch
git checkout -b security/fix-cve-YYYY-NNNNN

# Verify branch
git branch --show-current

For forks:

git clone git@github.com:<username>/agents-at-scale-ark.git
cd agents-at-scale-ark
git remote add upstream git@github.com:mckinsey/agents-at-scale-ark.git
git fetch upstream
git checkout -b security/fix-cve-YYYY-NNNNN upstream/main

Implementation

Applying the Fix

Once user approves and repository is cloned, apply changes:

cd agents-at-scale-ark

# For Go dependencies
go get package@v1.2.3
go mod tidy

# For Node.js dependencies
npm install package@1.2.3
npm audit fix

# For Python dependencies
# Edit requirements.txt or pyproject.toml
pip install -r requirements.txt

# For Docker base images
# Edit Dockerfile FROM statements

Verification

Basic Testing
cd agents-at-scale-ark

# Run tests
make test

# Build to check for breaking changes
make build

# Search for remaining vulnerable patterns
grep -r "vulnerable-pattern" .
Integration Testing with Setup Skill (Optional)

For changes that affect Ark runtime behavior, use the setup skill to test in a live cluster:

When to use setup skill for testing:

  • Go operator changes (controllers, webhooks, CRDs)
  • Service updates (ark-api, executor services)
  • Changes that affect Kubernetes interactions
  • Breaking changes that need verification

Setup skill workflow:

  1. Creates a Kind cluster
  2. Builds ark-cli from your security fix branch
  3. Installs Ark with your changes
  4. Verifies all pods are running
  5. Allows you to test the fix in action

Skip integration testing if:

  • Only updating documentation or CLI
  • Changes are in isolated utility functions
  • Dependencies don't affect runtime behavior

Security-Focused PR Templates

Commit Message Template

Ensure you're in the cloned repository:

cd agents-at-scale-ark
git add .

git commit -m "$(cat <<'EOF'
fix: CVE-YYYY-NNNNN in [component]

## Vulnerability Details
- CVE: CVE-YYYY-NNNNN
- Severity: [Critical/High/Medium/Low]
- CVSS Score: [X.X]
- Component: [package/library]

## Impact on Ark
[How this affects Ark services and realistic risk level]

## Changes
- Updated [component] from v[X] to v[Y]
- [Any code changes]

## Testing
- [Tests run and results]

## References
- CVE: https://cve.circl.lu/cve/CVE-YYYY-NNNNN
- Advisory: [URL]

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
EOF
)"

Push to Remote

# Push the security fix branch
git push origin security/fix-cve-YYYY-NNNNN

Pull Request Template

Create the PR with detailed security information:

gh pr create --title "fix: CVE-YYYY-NNNNN in [component]" --body "$(cat <<'EOF'
## Summary
Addresses security vulnerability CVE-YYYY-NNNNN in [component].

## Vulnerability Details

| Field | Value |
|-------|-------|
| **CVE** | CVE-YYYY-NNNNN |
| **Severity** | [Critical/High/Medium/Low] |
| **CVSS Score** | [X.X] |
| **Component** | [package] |
| **Current Version** | [old] |
| **Patched Version** | [new] |

### Description
[What the vulnerability is and how it could be exploited]

## Impact on Ark

### Affected Components
- [Service 1]: [Impact]
- [Service 2]: [Impact]

### Risk Assessment
**Risk Level**: [Level]

[Realistic assessment of actual risk to Ark deployments]

## Changes Made
- Updated `[component]` from `v[X]` to `v[Y]`
- [Other changes]

## Testing
- ✅ Unit tests pass
- ✅ Integration tests pass
- ✅ Manual verification completed

## References
- **CVE**: https://cve.circl.lu/cve/CVE-YYYY-NNNNN
- **Advisory**: [URL]
- **Patch Notes**: [URL]

🤖 Generated with [Claude Code](https://claude.com/claude-code)
EOF
)"

Important Notes

CVE API Usage

The CIRCL CVE API:

  • Endpoint: https://cve.circl.lu/api/cve/{CVE-ID}
  • Returns JSON with CVSS scores, references, affected versions
  • No authentication required
  • Fallback: Use web search if API is unreachable

Ark Security Context

When assessing risk:

  • Architecture: Kubernetes operator managing AI workloads
  • Components: Go operator, Python services, Node.js CLI
  • Deployment: Typically cluster-internal, trusted environments
  • Focus areas: CRD controllers, API services, executor services

Skill Composition

This skill provides CVE-specific tools. It works best when combined with:

  • research skill - For web searches, vendor advisories, evidence gathering
  • analysis skill - For cloning Ark repo (read-only) and examining codebase structure
  • setup skill - For integration testing in a live Ark cluster
  • architecture skill - For understanding service boundaries and impact

Complete workflow example:

  1. Research CVE (this skill + research skill)
  2. Analyze impact (this skill + analysis skill)
  3. Clone for development (this skill)
  4. Implement fix (this skill)
  5. Test integration (this skill + setup skill, if needed)
  6. Create PR (this skill)

User Approval is Mandatory

Never implement changes without explicit user approval. This ensures:

  • User understands security implications
  • Approach aligns with security policies
  • Testing strategy is appropriate
  • Breaking changes are acknowledged

Common Vulnerability Types

Go Dependencies

  • Check: go.mod, go.sum
  • Update: go get package@version && go mod tidy
  • Scan: go list -m all

Node.js Dependencies

  • Check: package.json, package-lock.json
  • Update: npm install package@version
  • Scan: npm audit
  • Transitive deps: A vulnerable package may appear at multiple versions across nested node_modules/. Check ALL lockfiles: find. -name "package-lock.json" -exec grep -l "<package>" {} \; jq -r '.packages | to_entries[] | select(.key | contains("<package>")) | "\(.key)\t\(.value.version)"' <lockfile>
  • npm overrides: Use "overrides" in package.json to force a patched version. But overrides are global — they affect ALL consumers of that package. If two consumers need incompatible major versions (e.g., express 4.x needs path-to-regexp ~0.1.x while router needs ^8.x), an override will break one of them.
  • Version conflict resolution: When a transitive dep has incompatible version requirements, consider upgrading the parent package instead. Example: upgrading express 4.x to 5.x unified path-to-regexp to a single 8.x version, resolving the conflict without overrides.

Python Dependencies

  • Check: requirements.txt, pyproject.toml
  • Update: Edit requirements files
  • Scan: pip-audit (if available)

Docker Base Images

  • Check: Dockerfile FROM statements
  • Update: Change base image version
  • Scan: docker scan or vulnerability databases

平台分布

Claude Code

27.67%
按下载量换算80

windsurf

24.04%
按下载量换算70

OpenCode

17.95%
按下载量换算52

Codex

11.79%
按下载量换算34

Antigravity

7.49%
按下载量换算22

Gemini CLI

3.55%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。

继续浏览同类 Skills