layoff-intel

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

• Third-party content exposure detected (high risk: 0.95). The skill’s runtime workflow explicitly performs “web search” and “research employee-generated signals” (e.g., Glassdoor/Blind/Reddit/LinkedIn) in Phase 1 and Phase 5, which ingests outsider-authored free text from public pages into the agent’s LLM context via the web-fetch/search/osint steps.