claude-design-critic

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow includes “Take in the target” via a URL (WebFetch + screenshot) or pasted code, and if the user supplies an outsider URL/page, the fetched page text can be ingested into the agent’s LLM context for critique (public web content = outsider free text).