hasdata

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

• Third-party content exposure detected (high risk: 0.85). HasData’s required runtime workflow can ingest outsider-authored free text when it uses the Web Scraping API (POST /scrape/web) or Scraper Jobs (e.g., crawler/contacts/platform jobs) to fetch and return scraped page content/markdown/transcripts from third-party websites into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill calls https://api.hasdata.com/scrape/web at runtime (and polls /scrapers/... which returns short-lived download URLs) to fetch arbitrary external pages (references[].link) and inject their markdown/json into the model's RAG context, so remote content fetched during runtime can directly influence prompts/responses.