vs-qa

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

• Third-party content exposure detected (high risk: 0.90). This skill instructs the agent to browse and ingest arbitrary web pages via agent-browser (e.g., Phase 3/4 goto + snapshotForAI and the "Get All Links" / form-fill snippets) so untrusted public site content or user-provided URLs will be read and used to drive testing and fix actions, allowing indirect prompt-injection from third-party pages.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill navigates to and snapshots the user-provided target URL (e.g., TARGET_URL / "https://yourapp.com/login") at runtime and logs/feeds page.snapshotForAI() (snap.full) into the agent flow, so external page content fetched during execution directly becomes model context and thus can control prompts.