data-pipeline

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

• Third-party content exposure detected (high risk: 1.00). The pipeline explicitly crawls and ingests public third‑party sources (see extraction.md and orchestration.md: crawl.py and src/collectors/ including web_collector.py collecting Charity Navigator, ProPublica, Candid, CauseIQ, and arbitrary websites) and that untrusted scraped content is fed into reconciliation, scoring, and LLM-driven narrative generation (process_baseline.py / process_rich.py), so external content can materially influence decisions and actions.