The Agent Skills Directory

[SAFE]: The skill implements strong scope discipline by explicitly forbidding modifications to application logic, state management, and backend structures. This restriction prevents the agent from making unauthorized functional changes to the codebase.
[SAFE]: The audit protocol requires a phased design plan and mandatory user approval before any implementation occurs. This human-in-the-loop requirement ensures that all visual changes are reviewed and authorized by the user.
[SAFE]: No security concerns such as credential harvesting, remote code execution, or network exfiltration were detected. The skill primarily performs analytical and documentation tasks based on local project files.
[SAFE]: While the skill ingests external project data (such as design tokens and PRDs), it uses this information for visual evaluation rather than execution, which effectively mitigates common indirect prompt injection risks.

design-audit