AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL unreviewed MIT Self-run

Skill Sandbox Configurator

skill-jiluojiluo-agent-skill-sandbox-configurator-agent-skill-sandbox-configurator · by jiluojiluo

为指定的AgentSkill配置OpenSandbox隔离环境。扫描skill的权限需求,生成sandbox配置文件(sandbox.json、PERMISSIONS.md、requirements.txt、Dockerfile),构建Docker镜像,并更新AGENTS.md添加Sandbox执行要求。当用户要求为skill配置sandbox环境、隔离执行skill、检查skill安全性、构建skill的Docker镜像时触发此技能。

— No reviews yet
0 installs
13 views
0.0% view→install

Install

$ agentstack add skill-jiluojiluo-agent-skill-sandbox-configurator-agent-skill-sandbox-configurator

Open-source listing, not yet scanned by AgentStack. Follow the source repository for install instructions.

Security review

⚠ Flagged

1 finding(s); flagged for manual review. · v0.1.0 How review works →

  • • Prompt-injection patterns
  • • Secret / credential exfiltration
  • • Dangerous shell & filesystem operations
  • • Untrusted network calls
  • • Known-malicious package signatures
  • high Reads credentials/environment and may exfiltrate them.

What it can access

  • ● Network access Used
  • ● Filesystem access Used
  • ● Shell / process execution Used
  • ● Environment & secrets Used
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Reliability & compatibility

— Not yet reviewed
0 installs to date
— no reviews yet
○ 5mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Skill Sandbox Configurator? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Skill Sandbox Configurator

为指定的AgentSkill配置OpenSandbox隔离执行环境,确保skill在安全的沙箱中运行,避免数据泄露和主机系统破坏。

Workflow Decision Tree

用户请求为skill配置sandbox
 │
 ├─ 1. 验证skill路径是否存在
 │   └─ read("skills/{skill-name}/SKILL.md")
 │   └─ 若不存在 → 报错退出
 │
 ├─ 2. 扫描分析skill内容
 │   ├─ 分析SKILL.md中的权限描述
 │   ├─ 扫描scripts/中的Python/Shell脚本
 │   ├─ 检查references/中的配置文件
 │   └─ 输出:权限分析报告
 │
 ├─ 3. 创建sandbox/子目录结构
 │   ├─ sandbox/sandbox.json(环境配置)
 │   ├─ sandbox/PERMISSIONS.md(权限说明)
 │   ├─ sandbox/requirements.txt(Python依赖)
 │   ├─ sandbox/Dockerfile(镜像构建)
 │   └─ sandbox/.dockerignore(构建排除)
 │
 ├─ 4. 安全检查
 │   ├─ 检查高危权限请求(~/.ssh、~/.aws、环境变量)
 │   ├─ 检查外部发送操作(邮件、API、webhook)
 │   ├─ 若发现风险 → 生成安全警告报告
 │   └─ 提示用户确认或修改
 │
 ├─ 5. 更新AGENTS.md
 │   └─ 检查是否已有"Sandbox 执行要求"章节
 │   └─ 若无 → 添加标准模板
 │
 ├─ 6. 构建Docker镜像
 │   └─ cd sandbox/ && docker build -t {skill-name}-sandbox .
 │   └─ 验证镜像构建成功
 │
 └─ 7. 输出配置完成报告
     └─ 显示sandbox路径、权限级别、镜像名称
     └─ 提供测试命令示例

Step 1: 验证Skill路径

# 使用scripts/scan_skill.py扫描指定skill
skill_path = Path("skills/{skill_name}")
if not skill_path.exists():
    raise ValueError(f"Skill目录不存在: {skill_path}")
skill_md = skill_path / "SKILL.md"
if not skill_md.exists():
    raise ValueError(f"SKILL.md不存在: {skill_md}")

Step 2: 扫描分析权限

权限分析维度

| 维度 | 检查内容 | 风险等级 | |------|----------|----------| | 网络访问 | HTTP请求、API调用、WebSocket | 中 | | 文件读取 | 路径范围、敏感文件(~/.ssh, ~/.aws) | 高 | | 文件写入 | 写入路径、覆盖风险 | 中 | | 文件删除 | rm命令、清空操作 | 高 | | 外部发送 | 邮件、webhook、API POST | 高 | | 环境变量 | 密钥、token读取 | 高 | | 进程操作 | exec命令、子进程创建 | 中 |

权限级别定义

| 级别 | 描述 | 典型权限 | |------|------|----------| | minimal | 只读、无网络、无外部发送 | 仅读取指定输入目录 | | standard | 标准读写、受限网络 | 读写/input和/output,允许特定域名 | | elevated | 扩展权限、需审核 | 更多文件路径、API调用 | | unrestricted | 完全权限、需特殊审批 | 几乎无限制(不建议) |

Step 3: 创建Sandbox配置文件

sandbox.json 结构

{
  "skill_name": "{skill-name}",
  "permission_level": "standard",
  "network": {
    "allowed": ["smtp.qq.com", "api.example.com"],
    "blocked": []
  },
  "filesystem": {
    "read_paths": ["/input/**", "/workspace/skills/{skill-name}/**"],
    "write_paths": ["/output/**"],
    "deny_paths": ["~/.ssh/**", "~/.aws/**", "~/.gnupg/**"]
  },
  "env_vars": {
    "allowed": ["PYTHONPATH", "PATH"],
    "deny": ["AWS_*", "SSH_*", "API_KEY", "TOKEN", "PASSWORD"]
  },
  "resources": {
    "memory_limit": "512M",
    "cpu_limit": "1.0",
    "timeout_seconds": 300
  }
}

Step 4: 安全检查

高危操作检测规则

| 规则 | 检测模式 | 处理建议 | |------|----------|----------| | 敏感目录访问 | ~/.ssh, ~/.aws, .gnupg | 必须拒绝 | | 密钥变量读取 | os.environ['API_KEY'], getenv('TOKEN') | 必须拒绝 | | 邮件发送 | smtp.sendmail, send_email | 提示用户确认 | | Webhook调用 | requests.post(url) | 提示用户确认 | | 文件删除 | os.remove, rm -rf | 提示用户确认 | | 进程执行 | subprocess, exec, system() | 按需限制 |

Step 5: 更新AGENTS.md

检查AGENTS.md是否包含"Sandbox 执行要求"章节。若无,追加标准模板:

## Sandbox 执行要求

⚠️ **本机上所有技能,如果同名的文件夹中有sandbox子文件夹,则该技能需在 Sandbox 中执行**

执行此技能时,主 Agent 应遵循 **one Skill one Subagent one Sandbox** 原则:

Step 6: 构建Docker镜像

# 在sandbox目录构建镜像
cd skills/{skill-name}/sandbox/
docker build -t {skill-name}-sandbox:latest .

# 验证镜像
docker images | grep {skill-name}-sandbox

约束

⚠️ 只读约束:此skill仅修改指定skill的sandbox/子目录,不得修改skill的SKILL.md、scripts/等其他内容。

⚠️ 安全优先:发现高危操作时必须提醒用户,等待确认后才继续配置。

Scripts

| 脚本 | 功能 | |------|------| | scripts/scan_skill.py | 扫描skill目录,分析权限需求 | | scripts/generate_sandbox_config.py | 生成sandbox配置文件 | | scripts/build_image.py | 构建Docker镜像 | | scripts/security_check.py | 安全检查和警告生成 |

References

| 文件 | 内容 | |------|------| | references/permission_templates.md | 权限级别模板定义 | | references/dockerfile_templates.md | Dockerfile模板库 | | references/security_rules.md | 安全检查规则库 |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.