低风险 — 风险评分 20/100
上次扫描:1 天前 重新扫描
20 /100
tray
Tray.io low-code automation platform integration via Membrane CLI
This is a legitimate Tray.io integration skill using the Membrane CLI. Shell execution is required for npm commands but undocumented; behavior otherwise matches stated purpose.
技能名称tray
分析耗时28.8s
引擎pi
可以安装
Add 'shell:WRITE' to the allowed tools declaration in SKILL.md metadata to match the actual requirements. Pin the CLI version (e.g., @membranehq/[email protected]) for reproducible installs.

安全发现 2 项

严重性 安全发现 位置
低危
Missing allowed-tools declaration 文档欺骗
SKILL.md metadata lacks allowedTools. The skill requires shell:WRITE (npm install, membrane CLI) and network:WRITE (API proxying), but neither is declared.
metadata section with empty allowedTools
→ Add allowedTools: [Bash] to the skill metadata. This is documentation hygiene rather than malicious隐瞒.
SKILL.md:1
低危
Unpinned npm package version 供应链
CLI is installed with 'npm install -g @membranehq/cli' without a version pin. Future versions could introduce unexpected behavior.
npm install -g @membranehq/cli
→ Pin to a specific version: npm install -g @membranehq/cli@latest or @membranehq/[email protected]
SKILL.md:47
资源类型声明权限推断权限状态证据
命令执行 NONE WRITE ✗ 越权 SKILL.md:47-48 npm install -g @membranehq/cli; SKILL.md:51 membrane login
网络访问 NONE WRITE ✗ 越权 SKILL.md:65 membrane request proxies API calls
文件系统 NONE NONE No file operations detected
环境变量 NONE NONE No environment variable access
数据库 NONE NONE No database access
剪贴板 NONE NONE No clipboard access
浏览器 NONE NONE Browser used only via system opener for auth flow, not automation
技能调用 NONE NONE No cross-skill invocation
2 项发现
🔗
中危 外部 URL 外部 URL
https://getmembrane.com
SKILL.md:7
🔗
中危 外部 URL 外部 URL
https://developers.tray.io/
SKILL.md:19

目录结构

1 文件 · 4.3 KB · 124 行
Markdown 1f · 124L
└─ 📝 SKILL.md Markdown 124L · 4.3 KB

依赖分析 1 项

包名版本来源已知漏洞备注
@membranehq/cli unpinned npm Version not pinned in SKILL.md install instructions

安全亮点

✓ Membrane CLI is from a known vendor (MembraneHQ) with an official GitHub repo
✓ No credential harvesting — Membrane manages auth server-side with no local secrets
✓ No data exfiltration — all API calls route through Membrane's documented proxy
✓ No obfuscation, base64, or suspicious encoding patterns
✓ No access to sensitive paths (~/.ssh, ~/.aws, .env)
✓ No reverse shell, C2, or outbound data transfer to untrusted endpoints
✓ Skill purpose (Tray.io automation) is clearly and accurately documented