career-jd-analyzer

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

• Third-party content exposure detected (high risk: 0.85). Yes—Step 1 “Intake & Fetch” uses WebFetch to fetch JD正文 from a user-provided link (runtime scraping of public job-posting pages), and that outsider-authored free-form text is then parsed/used as LLM context in Steps 2–4.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly performs a WebFetch at runtime to retrieve real JD pages (e.g. "https://www.zhipin.com/job_detail/...") which are then parsed and used to drive the agent's outputs, so these external job-platform URLs directly control the prompts/behavior at runtime.