brand-context

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md requires auto-drafting by scraping the brand’s website at runtime via firecrawl_urls_scrape / web_scrape_page, which ingests outsider-authored free text from public web pages into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill runtime scrapes user-supplied brand site URLs via the firecrawl_urls_scrape / web_scrape_page tools (e.g., the brand homepage URL such as https://brand.example.com) and injects the scraped page content into the drafting flow, which directly controls the agent's generated prompts/output.