playwright-stealth

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs the agent to ask the user for login credentials and to inject them into page.fill (and also suggests exporting clearance cookies), which requires the agent to receive and handle secrets and risks those secrets being echoed, logged, or otherwise exfiltrated by the LLM.

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow uses the user-supplied targetUrl and then ingests outsider-authored free text from the fetched page via page.goto(targetUrl, ...) followed by page.title() checks and page.evaluate(...).innerText/document metadata extraction.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill instructs modifying the host (apt-get/npm installs) and explicitly tells the agent to disable sandboxing (--no-sandbox), which are actions that change system state and bypass security protections.