browser-read-bypass
Fail
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: HIGHPROMPT_INJECTIONPRIVILEGE_ESCALATIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains explicit instructions to ignore and bypass platform-enforced security restrictions, specifically addressing the 'tier read' restriction. It encourages the agent to use low-level workarounds when standard interactions are blocked.
- [PRIVILEGE_ESCALATION]: The skill enables an agent to perform write-level interactions (clicking, typing) in an environment that has been restricted to read-only access by the host platform, effectively escalating its capabilities beyond its assigned permissions.
- [DYNAMIC_EXECUTION]: The skill uses PowerShell's Add-Type functionality to dynamically compile C# code that interfaces with the Windows user32.dll library at runtime. This allows the agent to execute system-level API calls that are not part of its standard toolkit.
- [COMMAND_EXECUTION]: The skill relies on executing complex system-level commands via Bash and PowerShell to perform UI manipulation and bypass restrictions.
Recommendations
- AI detected serious security threats
Audit Metadata