kelly-picks

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime ingest path reads into Busabase from outsider-authored web/content sources via the sweep workflow (configured browser_agent sources) and stores free-text fields like candidate name, summary, why_it_matters, and proposal brief/reason through node scripts/ingest_trends.mjs <payload.json>, which the LLM can then consume when rendering proposals/candidates from Busabase.