ads-dna

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches arbitrary public website content using the WebFetch tool ("Step 2: Fetch Pages" in SKILL.md) — returning visible text, blocks, meta tags and og:image — and then directly reads and interprets that untrusted, user-provided webpage content to build brand-profile.json and drive downstream /ads and visual-designer behavior, which enables indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill uses the WebFetch tool to retrieve the user-supplied website at runtime (e.g., https://yoursite.com), and the fetched HTML/CSS/meta/og-image content is injected into the agent's extraction/generation logic, so external page content directly controls the model's prompts and outputs.