---
title: "ClawGuard: Runtime Security for Tool-Augmented Agents"
created: 2026-04-15
updated: 2026-04-15
type: concept
tags: [agent, security, runtime, tool-use, indirect-prompt]
sources: [raw/papers/2026/04/2604.11790.md]
---
# ClawGuard: Tool-Augmented Agent 运行时安全

## 核心
针对 tool-augmented LLM agent 的间接 prompt 注入攻击，提供运行时安全框架。

## 与 Wiki 的关联
- [[agent-supply-chain-attack]] 关注 API router 层
- ClawGuard 关注 tool 调用层的运行时防护
- 两者互补构成 agent 安全的全链路视角

## Related
- [[agent-supply-chain-attack]]
- [[saver-faithful-reasoning]]