visual-debugger

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt explicitly shows and instructs filling credentials via browser actions (e.g., mcp__playwright__browser_type with plaintext password examples) and tells the agent to ask the user for test credentials or embed them, which would require the LLM to include secret values verbatim in tool calls or generated output.

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

• Third-party content exposure detected (high risk: 1.00). The skill navigates to and inspects arbitrary web pages (e.g., using mcp__playwright__browser_navigate with user-provided URLs) and then reads DOM/console/network data via browser_evaluate, browser_snapshot, and browser_console_messages, which clearly ingests untrusted third‑party web content.