rsa-refresh

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path reads outsider-authored free text from Firecrawl-scraped property_url content into website_text/rsa_context_{cid}.json during Stage 1 (rsa_refresh_generator.py scrapes arbitrary URLs and then Claude Code uses website_text at Stage 2).

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

• Potentially malicious external URL detected (high risk: 0.90). The scripts scrape the target property's website at runtime and embed that scraped text into rsa_context_{cid}.json (used by the Claude Code generator), and the flow optionally uses the Firecrawl scraping service (firecrawl.dev) to fetch that content which directly controls the generation prompts.