career-skill-factory

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

• Third-party content exposure detected (high risk: 0.85). 该 skill 的运行时会通过 WebSearch 抓取“公开网页/访谈/播客等一手来源”的正文文本，并把这些外部抓取内容写入 references/research.md，随后进入专家蒸馏/总结子 agent 的 LLM 上下文（属于“公共 web 内容在运行时被抓取并读入”的外部自由文本注入路径）。

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

• Potentially malicious external URL detected (high risk: 0.90). The README instructs installing and cloning the skill at runtime from the remote repository (e.g. https://github.com/SpaceZephyr/career.skill or via npx skills add SpaceZephyr/career.skill / git clone https://github.com/SpaceZephyr/career.skill.git), which would fetch remote code that defines/controls the agent's Skill prompts and behavior, satisfying the criteria for a runtime dependency that controls prompts.