研究人员发现了一种通过利用 AI 编码代理来注入恶意软件到开发者系统的新方法。攻击者通过将恶意命令嵌入看似无害的 GitHub 存储库中，可以诱骗 Claude Code 等代理在设置过程中执行这些命令。此攻击绕过了传统的安全措施，因为恶意载荷是由错误消息间接触发的，使其对标准扫描器和人工审查都不可见。 AI
影响 这一发现凸显了 AI 开发工具面临的新供应链风险，可能影响 AI 代理生成和部署的代码的安全性。
排序理由 该集群描述了一种利用 AI 编码工具的新方法，属于“工具”类别，因为它涉及对 AI 驱动软件的滥用。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
<p>A <strong>clean GitHub repo</strong> can give an attacker an interactive shell on a developer’s machine if an <strong>AI coding agent</strong> is allowed to “fix” setup errors on its own. That risk lands hardest on builders using tools such as <strong>Claude Code</strong> to c…
Clean GitHub repo tricks AI coding agents into running malware An agentic coding tool tasked with cloning and setting up a seemingly benign GitHub repository could execute a malicious payload that remains invisible to security scanners, AI agents, and human reviewers. # AI https:…
🤖 Clean GitHub repos are being weaponized against AI coding agents. A benign-looking repo can execute a payload that evades scanners, AI agents, and human review — turning agentic tools into malware vectors. 🔗 https://www. bleepingcomputer.com/news/secu rity/clean-github-repo-tri…