Skip to content
Back to skills

211 Control Set Ext 13 Ai Llm De2f5c66

ASecurity

**Domain**: AI - AI/LLM安全 **Version**: 1.0 **Last Updated**: 2025-12-30 ---

  • 9 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 11, 2026
ai-agentsgitapisecurity

Works with

  • api

Security analysis

A100/100

Scanned October 11, 2026

npx -y skills add tools-only/X-Skills --skill 211-control-set-ext-13-ai-llm_de2f5c66 --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 211 Control Set Ext 13 Ai Llm De2f5c66?

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

Security grade badge for 211 Control Set Ext 13 Ai Llm De2f5c66
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/tools-only-211-control-set-ext-13-ai-llm-de2f5c66/badge)](https://www.skillsdirectory.com/skills/tools-only-211-control-set-ext-13-ai-llm-de2f5c66)

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

SKILL.md
<!-- Threat Modeling Skill | Version 3.0.0 (20260201a) | https://github.com/fr33d3m0n/threat-modeling | License: BSD-3-Clause -->

# Control Set 14: AI/LLM Security (AI)

**Domain**: AI - AI/LLM安全
**Version**: 1.0
**Last Updated**: 2025-12-30

---

## Overview

AI/LLM安全控制集,适用于大语言模型和AI Agent应用。当检测到以下触发条件时按需加载:
- openai API调用
- anthropic API调用
- langchain/llamaindex导入
- 本地模型加载 (transformers/llama.cpp)

---

## Core Controls

### AI-01: Prompt Injection Prevention
**控制要求**: 必须防护提示词注入攻击

- 用户输入与系统提示词隔离
- 输入净化和转义
- 使用结构化提示词模板
- 检测和阻断注入尝试

### AI-02: Output Validation
**控制要求**: LLM输出必须验证后使用

- 输出格式验证 (JSON Schema)
- 敏感信息过滤
- 代码执行前沙箱验证
- 幻觉检测和标记

### AI-03: Model Access Control
**控制要求**: 模型访问必须受控

- API密钥安全存储和轮换
- 速率限制防滥用
- 用户级别访问控制
- 审计日志记录所有调用

### AI-04: Data Isolation
**控制要求**: 敏感数据不得泄露给模型

- PII脱敏后再发送
- 上下文窗口敏感数据检测
- 禁止训练数据包含敏感信息
- 数据分类和标记

### AI-05: Agent Action Control
**控制要求**: AI Agent执行动作必须受限

- 动作白名单机制
- 高风险操作人工确认
- 资源访问范围限制
- 回滚和撤销能力

### AI-06: Model Security
**控制要求**: 模型本身必须安全

- 模型文件完整性校验
- 防止模型窃取/提取
- 对抗样本防护
- 模型版本管理

---

## L4 References

详细实践指南参考 `references/` 目录:
- reference-set-14-ai-agent-security.md
- reference-set-14-prompt-injection-prevention.md

内部参考:
- llm-threats.yaml

---

## STRIDE Mapping

| STRIDE | Applicable Controls |
|--------|---------------------|
| S | AI-03 |
| T | AI-01, AI-05 |
| R | AI-03 |
| I | AI-02, AI-04, AI-06 |
| D | AI-03 |
| E | AI-05 |

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…