jobkorea-talent-search

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

• Third-party content exposure detected (high risk: 0.85). Runtime path: the skill fetches JobKorea’s search page/AJAX HTML (fetch() → post_search()), then converts scraped candidate row text into LLM context via clean_text()/parse_candidates() and finally prints it in the shortlist (print_markdown), so any free-form text present in JobKorea’s returned candidate summaries (outsider-authored by other users) can be ingested.