market-copy

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path includes python3 scripts/analyze_page.py <url> which fetches and extracts existing homepage copy from a user-supplied URL (public web content), so outsider-authored free text can be ingested into the LLM context for inline voice inference.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's pre-flight step instructs running "python3 scripts/analyze_page.py " to fetch the user-supplied homepage () at runtime and infer brand voice that directly shapes the agent's prompts/instructions, so external page content can control outputs.