create-brand-theme

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

• Third-party content exposure detected (high risk: 0.90). This skill's Step 1 "Path A — Website extraction" explicitly opens arbitrary public websites (via a browser MCP or WebFetch), scrapes CSS, fonts, images/SVGs, and writes photography/style summaries which are then baked into prompts and used to drive generate_image/render/update_theme calls, so untrusted third‑party page content can materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches user-provided website and asset URLs at runtime (e.g., fetching site content and asset URLs like "https://…/source.svg" or a brand site) to extract colors, fonts, and logos that are then baked into prompts and generation calls, so remote content directly controls agent prompts.