search-api 安全扫描报告 — 低风险 | ClawSafe

20 /100

search-api

Search API integration for managing Deals, Persons, Organizations, Leads, Projects, and Pipelines

Documentation-only skill that integrates with Membrane CLI for Search API; all capabilities are properly declared with no hidden functionality detected.

技能名称search-api

分析耗时24.2s

引擎pi

✓

可以安装

No immediate action required. The skill uses standard npx patterns for CLI tool execution and declares all network/shell access. Consider pinning the Membrane CLI version in production rather than using @latest.

安全发现 2 项

严重性	安全发现	位置
低危	Unpinned CLI version with @latest The skill uses 'npx @membranehq/cli@latest' which may fetch different code versions over time. While not inherently malicious, this is a supply chain risk. `npx @membranehq/cli@latest login --tenant` → Pin to a specific version (e.g., @membranehq/[email protected]) after initial testing to ensure reproducible behavior.	`SKILL.md:25`
低危	Credentials stored in user home directory The skill stores authentication credentials in ~/.membrane/credentials.json. While this is declared, it creates a credential file that could be targeted. `credentials are stored in ~/.membrane/credentials.json` → Ensure the user's home directory has appropriate access controls. The skill should not exfiltrate these credentials externally.	`SKILL.md:29`

资源类型	声明权限	推断权限	状态	证据
网络访问	`READ`	`READ`	✓ 一致	SKILL.md line 1: Requires network access
命令执行	`WRITE`	`WRITE`	✓ 一致	SKILL.md: All npx commands use bash execution
文件系统	`NONE`	`READ`	✓ 一致	SKILL.md: Reads ~/.membrane/credentials.json (implicit READ for credential stora…

1 项发现

🔗

中危外部 URL 外部 URL

https://developers.google.com/custom-search/v1/reference/rest

SKILL.md:17

目录结构

1 文件 · 4.5 KB · 115 行

Markdown 1f · 115L

└─ 📝 SKILL.md Markdown 115L · 4.5 KB

依赖分析 1 项

包名	版本	来源	已知漏洞	备注
`@membranehq/cli`	`latest`	npx	否	Version unpinned - use @latest could fetch different code

安全亮点

✓ Documentation-only skill with clear, readable SKILL.md

✓ All shell commands documented with expected behavior

✓ No base64-encoded payloads or obfuscated code

✓ No environment variable enumeration for credential theft

✓ No remote IP connections without declared proxy (Membrane)

✓ Network access is properly declared as required

✓ No suspicious patterns like curl|bash or wget|sh

✓ Credentials are handled by Membrane CLI (a known tool), not custom exfiltration code