MCP Implementation Security Review
Use when reviewing an MCP server, client, or tool handler for authentication, sessions, rate limits, schema validation, RCE vectors, and OWASP MCP risks before release.
Install
npx skills add sandbaseai/workbuddy-skill --skill mcp-implementation-security-reviewgit clone --depth 1 https://github.com/sandbaseai/workbuddy-skill.git /tmp/workbuddy-skill
mkdir -p ~/.claude/skills
cp -r /tmp/workbuddy-skill/skills/mcp-implementation-security-review ~/.claude/skills/mcp-implementation-security-reviewPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# MCP 实现安全评审 对 MCP Server、Client 和工具处理器做可复核的安全评审,重点覆盖网络暴露、身份隔离、会话、限流、输入 Schema、RCE、供应链和 OWASP MCP 风险。输出文件/行号证据、反证、未知项和修复优先级,不把静态扫描或分数当成安全认证。 ## 评审边界与目标分类 - 开始前锁定目标仓库、commit、MCP 协议版本、Server/Client 类型、传输方式、部署拓扑、工具清单、数据范围、owner 和评审授权。 - 默认只读检查源码、配置、依赖锁文件、服务清单、审计样例和文档;不得启动未知服务、向生产端点发包、调用真实工具、修改权限或执行未审查脚本。 - 先判断网络暴露:HTTP/SSE、Streamable HTTP、监听 `0.0.0.0`、容器端口映射或反向代理都按远程服务处理;STDIO 只是不适用网络基线,不代表安全。 - 文档仅在描述目标自身的 Server 行为、部署或认证时作为证据;模板、教程、vendor、node_modules 和客户端配置不能冒充服务实现证据。 - 任何无法从声明范围验证的控制标记为 `NEEDS INVESTIGATION`,并列出需要的文件、配置、运行测试或 owner,不猜测为通过。 ## 远程 MCP 基线 对网络暴露的 Server 逐项检查;对本地 STDIO 给出最佳实践并继续执行 RCE 检查。 ### MCP-01 身份隔离 确认每个 MCP 路由都验证可信身份并执行授权,不能把 session ID、网络位置或上一次请求当认证。Server 应有独立的 audience/resource 和应用身份,出站调用使用独立的最小权限凭据;OAuth/MCP discovery 端点只能返回元数据,不能执行工具或暴露数据。 ### MCP-02 会话安全 若没有 session ID,记录 `N/A`,但仍要求每请求认证。若使用会话,确认每个请求重新认证授权,ID 是不可预测的 CSPRNG 关联令牌,绑定已认证上下文、不编码权限、不出现在 URL,丢失或复用 ID 不能获得权限。SDK 自动生成但源码不可见时标为 `NEEDS INVESTIGATION`。 ### MCP-03 限流与滥用防护 工具发现和调用必须在 MCP Server 运行时限流,而不只依赖网关;按身份、会话和工具成本分桶,对写入/高成本工具更严格。超限须在后端动作前 fail-closed,返回 429 与 `Retry-After`。记录窗口、突发量、并发、成本、租户隔离和恢复行为。 ### MCP-04 工具 Schema 校验 每次调用都在 Server 边界依据显式 Schema 校验类型、必填项、枚举、长度/数值边界和未知属性;默认拒绝额外属性。非法输入返回 400/MCP 错误且不触发后端动作。客户端或下游的校验不能替代服务端校验,必须有负向测试。 ### MCP-05 官方 SDK 与维护 确认远程 Server 使用目标语言官方 MCP SDK 或提供同等控制的直接证据,记录协议版本、依赖锁定、补丁状态和 SDK 自动覆盖/未覆盖的控制。手写 HTTP/SSE 栈不能自动视为安全;不确定时标 `NEEDS INVESTIGATION`。 ## RCE 与数据出口检查 逐个工具追踪不可信参数到执行边界,并将每项标为 `SAFE`、`AT RISK` 或 `N/A`:
Details
- Skill name
mcp-implementation-security-review- Repository
- sandbaseai/workbuddy-skill
- Path
skills/mcp-implementation-security-review/SKILL.md- License
- MIT
- Repo updated
- 8 September 2026
FAQ
What does the MCP Implementation Security Review skill do?
Use when reviewing an MCP server, client, or tool handler for authentication, sessions, rate limits, schema validation, RCE vectors, and OWASP MCP risks before release.
How do I install the MCP Implementation Security Review skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add sandbaseai/workbuddy-skill --skill mcp-implementation-security-review. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is MCP Implementation Security Review popular?
It lives in sandbaseai/workbuddy-skill, which has 9 GitHub stars and 3 forks.
Related skills
Access Control Security Best Practices
9sandbaseai/workbuddy-skill
Design, review, implement, and verify least-privilege access controls across users, services, tenants, resources, and administrative workflows.
Accessibility Review
9sandbaseai/workbuddy-skill
Use when reviewing a web interface or interaction for semantic, keyboard, focus, screen-reader, ARIA, and contrast accessibility risks.
Acquire Codebase Knowledge
9sandbaseai/workbuddy-skill
Use when the user asks to map, document, or onboard into an existing codebase and needs evidence-backed architecture, structure, conventions, integrations, testing, and concerns documentation.
MCP Release QA
9sandbaseai/workbuddy-skill
Use when reviewing or releasing an MCP server, tool, resource, prompt, catalog, or installation path and runtime protocol evidence is required.
MCP CLI
9sandbaseai/workbuddy-skill
Use when discovering, inspecting, or calling MCP servers and tools from a command line, with explicit schema, target, permission, and side-effect checks.
MCP Security
9sandbaseai/workbuddy-skill
Use when designing, reviewing, or operating MCP servers, tool registries, or multi-agent pipelines; establish trust boundaries, validate tool inputs and outputs, propagate authorization, and contain prompt-injection and data-exfiltration risk.
Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.