Vague Triggers
Medium
- Confidence
- 89% confidence
- Finding
- The skill metadata explicitly allows opening arbitrary websites from flexible user input such as names, partial URLs, and full URLs, but it does not describe any confirmation step, allowlist, or safety boundary. In an agent setting, this can be abused for phishing, drive-by navigation, or surprising browser actions triggered from ambiguous prompts, making the broad behavior a real security concern even though it appears intended as convenience functionality.
