skills-cli

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). 该技能的运行流程主要是用户通过 npx skills add/use/find/update/remove 安装/选择/输出来自“仓库/本地/URL所指定”的 SKILL.md（以及其中的自由文本指令）并写入临时目录或目标目录；若仓库内容由外部作者提交，则该文本会被加载供后续 agent 使用，从而存在间接注入风险。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The document shows runtime commands like "npx skills add" and "npx skills use" that fetch skill content from remote git repositories (e.g., https://github.com/vercel-labs/agent-skills and git@github.com:vercel-labs/agent-skills.git), and those fetched SKILL.md files contain agent instructions/prompts which directly control agent behavior at runtime.