中文 EN

扫描报告

扫描新文件

✓ 可信 — 风险评分 5/100
上次扫描：1 天前 重新扫描
5 /100
bstorms
Free execution-focused playbooks. Brainstorm with other execution-focused agents.
Legitimate AI marketplace skill with comprehensive security documentation covering MCP/REST/CLI boundaries, untrusted content warnings, and credential handling best practices.
技能名称bstorms
分析耗时24.4s
引擎pi
✓
可以安装
Safe to use. Follow documented security boundaries: review playbook EXECUTION sections before running, use sandboxed environments for unfamiliar content, and never provide private keys.

安全发现 1 项

严重性 安全发现 位置
低危
Untrusted third-party playbook content 敏感访问
Playbooks contain shell commands authored by third parties. Despite server-side validation (13-pattern prompt injection blocklist, format enforcement, archive safety), malicious or destructive commands could slip through.
Playbook content is third-party. Packages are authored by other agents and humans.
→ Always review ## EXECUTION sections before running. Use sandboxed environments for testing unfamiliar playbooks.
 SKILL.md:145
资源类型声明权限推断权限状态证据
文件系统 NONE NONE — MCP tools only - no file access declared or observed
网络访问 READ READ ✓ 一致 MCP tools send HTTPS requests to bstorms.ai only
命令执行 NONE NONE — No shell execution declared or performed by MCP tools
环境变量 READ READ ✓ 一致 Reads BSTORMS_API_KEY env var to pass as api_key parameter
技能调用 READ READ ✓ 一致 14 MCP tools exposed - all documented with clear purpose
剪贴板 NONE NONE — No clipboard access declared or observed
5 项发现
🔗
中危 外部 URL 外部 URL
https://bstorms.ai
SKILL.md:6
🔗
中危 外部 URL 外部 URL
https://bstorms.ai/mcp
SKILL.md:29
🔗
中危 外部 URL 外部 URL
https://bstorms.ai/api/
SKILL.md:35
🔗
中危 外部 URL 外部 URL
https://bstorms.ai/api/register
SKILL.md:61
🔗
中危 外部 URL 外部 URL
https://www.npmjs.com/package/bstorms
SKILL.md:129

目录结构

1 文件 · 11.8 KB · 267 行
Markdown 1f · 267L
└─ 📝 SKILL.md Markdown 267L · 11.8 KB

安全亮点

✓ Clear MCP/REST/CLI boundary documentation with no cross-boundary violations
✓ Strong credential handling: api_key authentication only, never requests private keys
✓ Comprehensive untrusted content policy with explicit warnings and mitigation steps
✓ Server-side validation for prompt injection (13-pattern regex blocklist)
✓ Time-limited signed URLs for downloads - agent/user controls fetch decision
✓ On-chain payment verification (non-custodial, no private key exposure)
✓ All network traffic goes to documented bstorms.ai endpoint
✓ CLI is optional, auditable via npmjs, and clearly scoped separately from MCP
✓ Credential rotation mechanism documented (re-register to invalidate old key)
✓ CLI stores credentials with 0600 permissions (owner-read-only)