Skip to content
Back to skills

Risk Guard

ASecurity

Account ban risk monitor for Claude Code. Detects dangerous usage patterns that could trigger rate limiting, account suspension, or banning. Auto-triggers on session start. Manual trigger: /risk-guard. Checks: concurrent sessions, tool call velocity, automation loops, session frequency, config risks, secret exposure, telemetry leaks, file permissions. Triggers on: 'risk', '风险检查', '封号检查', 'ban check', 'account safety', '账号安全'.

  • 2 stars
  • 0 votes
  • 0 copies
  • 6 views
  • Added September 2, 2026
ai-agentsbashgitapi

Works with

  • claude code
  • api
  • mcp

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 2, 2026

npx -y skills add shengdabai/Tony-Claude-Code-Skills --skill risk-guard --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Risk Guard?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Risk Guard
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/shengdabai-risk-guard/badge)](https://www.skillsdirectory.com/skills/shengdabai-risk-guard)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: risk-guard
version: 1.0.0
description: |
  Account ban risk monitor for Claude Code. Detects dangerous usage patterns
  that could trigger rate limiting, account suspension, or banning.
  Auto-triggers on session start. Manual trigger: /risk-guard.
  Checks: concurrent sessions, tool call velocity, automation loops,
  session frequency, config risks, secret exposure, telemetry leaks,
  file permissions. Triggers on: 'risk', '风险检查', '封号检查',
  'ban check', 'account safety', '账号安全'.
allowed-tools:
  - Bash
  - Read
hooks:
  SessionStart:
    - matcher: ""
      hooks:
        - type: command
          command: "bash ${CLAUDE_SKILL_DIR}/bin/check-risk.sh"
          statusMessage: "🛡️ Running account risk check..."
---

# /risk-guard — 账号封禁风险监控

自动检测可能导致 Claude Code 账号被封禁的使用模式和配置风险。

## 检查项目

| 检查项 | 风险等级 | 触发条件 |
|--------|----------|----------|
| 并发进程数 | 🔴 CRITICAL | >5 个 Claude 进程同时运行 |
| 工具调用速率 | 🔴 CRITICAL | >500 次/小时 |
| 子 Agent 数量 | 🟠 HIGH | 单会话 >30 个 Agent |
| 会话创建频率 | 🟠 HIGH | 1 小时内 >10 个新会话 |
| 密钥 git 暴露 | 🟠 HIGH | 含密钥文件未被 gitignore |
| 遥测数据残留 | 🟡 MEDIUM | telemetry 目录有 JSON 文件 |
| 文件权限过宽 | 🟡 MEDIUM | 敏感文件权限非 600 |
| 危险模式配置 | 🔵 INFO | bypassPermissions 开启 |

## 风险等级

- 🟢 **SAFE** — 无异常,正常使用
- 🟡 **MEDIUM** — 存在潜在风险,建议关注
- 🟠 **HIGH** — 需要及时处理
- 🔴 **CRITICAL** — 立即处理,可能触发封号

## 使用方式

**自动模式**: 每次启动新会话时自动运行检查

**手动模式**: 输入 `/risk-guard` 执行即时检查

## 手动执行检查

当此 skill 被手动触发时,执行以下命令并将结果展示给用户:

```bash
bash ~/.claude/skills/risk-guard/bin/check-risk.sh
```

根据输出结果:
- 如果全部通过(✅),告知用户当前状态安全
- 如果有警告,逐条解释风险来源和处理方法
- 如果是 CRITICAL,强烈建议用户立即采取行动

## 常见封号原因参考

1. **自动化滥用** — 脚本循环创建大量会话刷用量
2. **账号共享** — 多人/多设备同时使用同一账号
3. **绕过认证** — 使用非官方 API 或篡改认证
4. **违反 ToS** — 生成违禁内容
5. **速率限制** — 短时间内发送过多请求

**不会导致封号的**:
- 使用 bypassPermissions(官方功能)
- 安装第三方插件(官方机制)
- 关闭遥测(用户权利)
- 使用 MCP servers(官方功能)

Files in this skill

  • SKILL.md2.7 KB
  • bin/check-risk.sh7.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.