html-to-bricks

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

• Third-party content exposure detected (medium risk: 0.30). In Step 4 (Mode A pasted in the conversation) and Step 4 (Mode B URL fetch) the workflow accepts outsider-authored HTML/CSS (including inline <style> and external stylesheet text) and then Step 5 calls respira_convert_html_to_builder with that HTML/CSS at runtime, so the LLM ingests attacker-controlled free text for conversion.