漏洞描述
Claude Code subagent configuration files under `.claude/agents/` were detected on the web server. Each `.md` file in this directory defines a specialized AI subagent (name, description, tool allow-list, system prompt) that Claude Code will delegate work to. Exposure of these files can leak custom system prompts (often containing internal API endpoints, business logic, prompt-engineering secrets), tool permission grants, and information about the project's internal automation setup. Directory listings on `.claude/agents/` further expose the full inventory of subagents defined for a project.
影响产品
暂无结构化产品信息。
修复建议
建议关注厂商安全公告，及时升级至已修复版本，并结合实际资产暴露情况采取缓解措施。