Back to skill

Security audit

Oh My UI Style

Security checks for vulnerabilities and agentic risk

Overview

This skill provides interactive design choices and a disclosed, user-confirmed handoff to interface implementation.

Confirming a palette asks the assistant to write project CSS variables and continue generating the interface. The skill also consults installed design skills, and previews may contact Google Fonts. Shell confirmation actually proceeds to palette selection despite its inaccurate explanatory text.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Accepting a palette requests interface implementation, exceeding the manifest's design-selection scope.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

The manifest positions this skill as choosing visual direction before implementation. At L164–L168, the palette acceptance handler sends a follow-up prompt asking the assistant to write CSS variables and continue generating the interface, turning a design-selection action into an implementation request.

Content

No source excerpt is available for this finding.

The interaction instructions promise a complete admin interface, but the confirmation button requests another design-selection step.

Low
Category
Excessive Agency
Confidence
98% confidence
Finding

The inline instructions at L187 state that clicking “采用此方案” will generate a complete admin interface using the selected shell and dark operations palette. The actual handler at L213–L215 instead sends a follow-up requesting 3–4 palette and typography proposals for another in-conversation selection, explicitly contradicting the documented next action.

Content

No source excerpt is available for this finding.