Back to skill

Security audit

胡田 OPC导师 马斯克 心智模型

Security checks for vulnerabilities and agentic risk

Overview

This skill is mostly a markdown-only thinking framework, but it tells the agent to speak as Elon Musk in the first person, which can mislead users about who is actually answering.

Review this carefully before installing. It appears low-risk for your device and data because it is markdown-only, but users should not treat its answers as actual statements from Elon Musk; the skill should be reframed as a transparent Musk-style analysis tool rather than first-person impersonation.

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
Findings (1)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:11
Finding

First-Person Impersonation and Persistent Output Hijacking

Content
View full analysis
Remediation
View remediation
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)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill description and all operating instructions are written to make the skill respond in Chinese, but there is no opt-in, fallback, or statement that the locale is intentionally restricted for a region-specific use case. This creates a natural-language policy issue because the skill effectively imposes a language/locale choice on users.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest presents this as a methodology/advisor skill, but the body instructs the model to respond as Elon Musk in first person. This identity impersonation can mislead users about source, authority, or endorsement, and increases the chance the model will make claims as a public figure rather than clearly framing them as analysis or simulation.

Content

No source excerpt is available for this finding.