ikevss-zh-friendly

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). Several URLs point to personal GitHub repositories and raw GitHub-hosted content that the skill instructs users to clone/download (unvetted third‑party code that can deliver arbitrary files), which is a higher-risk distribution channel for malware compared with official vendor/CDN sources.

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

• Third-party content exposure detected (high risk: 0.75). 该技能在运行时会从用户本机已安装的各 AI 工具的 Markdown 文件（例如 ~/.claude/skills/*/SKILL.md、~/.claude/commands/*.md、~/.claude/agents/*.md 及类似路径）读取技能 description/正文并将其内容喂给 LLM 进行“汉化/生成”，其中这些文本的作者并不一定是操作用户，从而构成“本机已装的第三方 skills/agents 文件（outsider-authored）→ 读取后进入 LLM 上下文”的间接提示注入风险。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runtime will fetch agent Markdown from the agency-agents-zh repository (e.g., https://raw.githubusercontent.com/ikevss/agency-agents-zh/main/{path}) to install agent personae whose frontmatter and System Prompt control agent behavior, so remote content directly influences prompts during execution.