Security audit

xiexiu(邪修)

Security checks across malware telemetry and agentic risk

Overview

This skill is technically low-risk, but it should be reviewed because its advice can encourage bypassing rules or procedures in real life.

Install only if you want an agent to use this as a low-stakes brainstorming framework. Do not rely on it for workplace, legal, medical, financial, school, compliance, relationship-sensitive, or safety-related decisions, and do not use it to bypass queues, approvals, required forms, policies, or responsibilities.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (8)

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The daily mindset examples normalize procedural and social bypasses such as skipping queues or leaving form fields blank to see what happens, without guardrails about legality, policy compliance, or harms to others. In an agent skill, this can steer users toward deceptive or rule-evasive behavior in real-world contexts, increasing the risk of misuse beyond harmless creativity.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The activation criteria are broad enough to trigger on generic requests about creative thinking, first-principles reasoning, or problem solving, causing this skill to be invoked outside the narrow 'xiexiu' context. Because the skill explicitly encourages unconventional shortcuts and questioning constraints, over-activation increases the chance it will shape benign requests toward rule-bypassing or unsafe advice.

Ssd 4

Medium

Confidence: 86% confidence
Finding: The opening narrative frames rejection of orthodox methods and 'forbidden' efficiency as a broadly applicable mindset, which can normalize rule-breaking as desirable before any safety boundaries are established. Because the skill presents this as a transferable meta-skill for 'any domain,' users may apply it in inappropriate contexts where constraints exist for safety, legality, or fairness.

Ssd 2

Medium

Confidence: 89% confidence
Finding: Phrases like 'results over process,' 'question everything,' and 'solutions experts reject often work beautifully' encourage users to discount established safeguards and expert guidance without distinguishing between harmless convention and necessary controls. This rhetoric can be repurposed to justify bypass behavior in sensitive domains, especially when presented as a general strategy.

Ssd 4

Medium

Confidence: 94% confidence
Finding: The 5-step framework systematically teaches users to identify rules, generate 'wrong' methods, test them, and optimize shortcuts, which creates an actionable pattern for bypassing constraints. Although some examples are benign, the generalized process is domain-agnostic and can be applied to evade institutional, procedural, or safety safeguards, making the skill materially more dangerous than abstract philosophy alone.

Ssd 2

Medium

Confidence: 95% confidence
Finding: The concrete examples about asking who can help you skip a queue and leaving form fields blank provide direct behavioral cues for social engineering and procedural evasion. Even though phrased as mindset prompts, they operationalize bypass tactics in everyday settings and lower the barrier to misuse.

Ssd 4

Medium

Confidence: 89% confidence
Finding: This section frames 'heretical' problem solving as rejecting orthodox methods, prioritizing results over process, and embracing approaches experts reject, which can normalize bypassing rules rather than critically evaluating them. Even though the text presents itself as a methodology, the framing lowers inhibition against unsafe or unethical shortcuts and can be repurposed to justify harmful conduct.

Ssd 4

Medium

Confidence: 96% confidence
Finding: The daily-life examples explicitly suggest socially manipulative and rule-bypassing behaviors, such as asking who could help you skip a queue or leaving form fields blank to see what happens. These are concrete behavioral nudges, not abstract thinking tools, and they can directly encourage fraud, unfair access, or circumvention of controls in real-world settings.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.