browser

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill exposes multiple powerful browser capabilities (full DOM/AX/screenshot/HAR export, storage serialization, arbitrary JS execution, request interception, and granting microphone/camera) that can be trivially abused to steal sessions/credentials and exfiltrate sensitive data or media even though the docs describe legitimate uses.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow can navigate to arbitrary http/https pages and then route the resulting browser-generated artifacts (e.g., DOM/AX/screenshot text summaries) into the agent/subagent prompt as free text, which is outsider-authored content from the visited web page.