Security audit

姜胡说思维教练（赚钱·自媒体·决策）

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese-language persona and analysis skill with disclosed roleplay, research, and sourcing expectations, and I found no executable code, persistence, credential handling, or destructive behavior.

Install this if you want a Chinese persona-style thinking advisor and are comfortable with it browsing or checking public facts for current topics. Treat its first-person answers as roleplay based on public material, not statements from the real person, and avoid relying on it for personalized financial advice.

A.I.G

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

SkillSpector

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 (3)

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The skill forces first-person impersonation and a fixed persona-speaking style without clear user opt-in at answer time, which can mislead users about source, certainty, and whether statements are authentic or inferred. While mainly a safety and transparency issue rather than a classic exploit, mandatory impersonation can suppress normal clarifications and make fabricated or role-inferred claims appear authoritative.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The skill materially expands from a passive persona/perspective prompt into a mandatory agentic workflow that requires external fact-finding and tool use before answering certain questions. This increases the skill's authority and action surface, can cause unrequested browsing or data retrieval, and may override the host assistant's normal safety, privacy, or product constraints around when tools should be used.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The entire skill is written exclusively in Chinese and does not indicate that users may choose another language or locale. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation unless the constraint is explicitly justified as region-specific.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.