中文 EN
Start Review
Skill Trust Decision

cursor-agent

Cursor Agent skill 为 Cursor CLI 提供文档化封装，存在文档-能力声明不一致和供应链风险，但无恶意行为证据

Install decision first Source: ClawHub Scanned: Apr 11, 2026
Files 3
Artifacts 5
Violations 0
Findings 4

Why this conclusion was reached

1/4 dimensions flagged
Pass
Declared vs actual capability

Declared resources and inferred behavior are broadly aligned.

Block
Hidden execution and egress

1 high-risk artifacts or egress signals were extracted.

Pass
Attack chain and severe findings

There is no explicit malicious chain in the report.

Review
Dependencies and supply chain hygiene

Dependency information is incomplete, so supply-chain confidence stays limited.

What drove the risk score up

文档-行为声明不一致 +15

文档声称'No raw shell access by default'，但实际通过 exec 命令可执行任意 shell

供应链风险 +8

curl|cursor.com/install|bash 远程脚本执行存在 RCE 风险

云端代理外传 +5

agent -c 可将工作推送到 cursor.com/agents 云端

Most important evidence

Medium Doc Mismatch

文档声称与实际能力不符

文档 Security 部分声称 'No raw shell access by default'，但 skill 提供的 exec 命令模式允许执行任意 shell 命令

SKILL.md:215
修正安全声明，明确说明通过 exec 执行 shell 命令的能力范围
Medium Supply Chain

远程脚本执行风险

安装说明使用 curl|bash 管道从 cursor.com 下载并执行安装脚本，存在供应链攻击风险

SKILL.md:46
使用官方包管理器（brew、npm）或预下载脚本验证方式安装
Low Data Exfil

云端代理功能可能涉及数据外传

agent -c 命令将工作内容推送到 Cursor 云端 (cursor.com/agents)，代码和数据可能被第三方处理

SKILL.md:47
明确告知用户云端处理的隐私影响，提供 opt-out 选项
Low Priv Escalation

影子权限：Cursor CLI 隐式加载项目规则

CLI 自动加载 .cursor/rules、AGENTS.md、CLAUDE.md、mcp.json，可能执行未声明的配置

SKILL.md:103
在安全说明中列出所有隐式加载的配置源

Declared capability vs actual capability

Shell Pass
Declared WRITE
→
Inferred WRITE
SKILL.md:exec command pattern
Network Pass
Declared READ
→
Inferred WRITE
agent -c cloud handoff, SKILL.md:47
Filesystem Pass
Declared NONE
→
Inferred WRITE
Cursor CLI performs read/write by design
Environment Pass
Declared NONE
→
Inferred READ
Cursor CLI loads AGENTS.md, CLAUDE.md, mcp.json

Suspicious artifacts and egress

Critical Dangerous Command
curl https://cursor.com/install -fsS | bash

README.md:46

Medium External URL
https://marketplace.visualstudio.com/items?itemName=xiaoyaner.openclaw-node-vscode

README.md:20

Medium External URL
https://clawhub.ai/xiaoyaner0201/vscode-node

README.md:21

Medium External URL
https://cursor.com/install

README.md:46

Medium External URL
https://cursor.com/agents

README.md:110

Dependencies and supply chain

There are no structured dependency warnings.

File composition

3 files · 661 lines
Markdown 2 files · 650 linesJSON 1 files · 11 lines
Files of concern · 2
README.md Markdown · 325 lines
curl https://cursor.com/install -fsS | bash · https://marketplace.visualstudio.com/items?itemName=xiaoyaner.openclaw-node-vscode · https://clawhub.ai/xiaoyaner0201/vscode-node · https://cursor.com/install · https://cursor.com/agents
SKILL.md Markdown · 325 lines
文档声称与实际能力不符 · 远程脚本执行风险 · 云端代理功能可能涉及数据外传 · 影子权限：Cursor CLI 隐式加载项目规则
Other files · _meta.json

Security positives

提供沙箱模式支持 (agent --sandbox)
提供命令审批机制
Node 路径使用 Ed25519 设备身份验证
Node 路径限制路径在 workspace 内
文档结构清晰，声明了 PTY 需求等实现细节