Back to skill

Security audit

去 AI 味界面文案

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently provides UI copywriting guidance, with disclosed installation steps and no evidence of hidden or unrelated behavior.

Review the installer and repository revision before running the documented global, automatically confirmed installation. Review proposed copy changes and specify your product’s locale conventions when they differ from the skill’s defaults.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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
Findings (3)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied command performs software installation, not the declared UI-copy task. The -g and -y flags request global installation without interactive confirmation. Although the installed skill's name is related to the description, that does not establish its behavior, and the supplied chunk contains no copywriting or review functionality. No malicious behavior is established, but the installation capability is not disclosed by the declared purpose.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The command npx skills add yzfly/skills@anti-ai-copy -g -y can download and execute an unpinned version of the skills npm package with the invoking user's permissions. A compromised package release could therefore execute malicious code; the remote skill content is also not fixed to an immutable revision. This is ordinary installation documentation, not evidence of malicious intent or an MCP server, but its copy-and-run format creates a genuine supply-chain exposure if followed.

Content

No source excerpt is available for this finding.

The skill mandates 24-hour time formatting without a user-selected locale or a documented regional justification.

Low
Category
Output Handling
Confidence
88% confidence
Finding

The instruction explicitly requires times to use the 24-hour clock. Unlike a locale-aware recommendation, this unconditional formatting rule can override the user's or product's established locale conventions without opt-in.

Content

No source excerpt is available for this finding.