Security audit

fanqie-masterclass

Security checks across malware telemetry and agentic risk

Overview

This static writing-course skill does not run code, but it needs Review because it teaches risky publishing tactics including plagiarism-style imitation, fake engagement, and specific tax-form guidance.

Install only if you specifically want this Fanqie/online-fiction monetization curriculum. Treat its promotion, imitation, and tax sections cautiously: do not use fake engagement, do not copy protected story structures or mislabel derivative work as original, and verify payout or tax-form decisions with official platform/IRS guidance or a qualified professional.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (8)

Description-Behavior Mismatch

Medium

Confidence: 92% confidence
Finding: This chapter materially expands from fiction-writing instruction into audience manipulation, traffic capture, private-domain funneling, and monetization playbooks across multiple platforms. In the context of a writing-teaching skill, these tactics encourage behavior that can violate platform rules, mislead audiences, or facilitate spammy growth schemes rather than simply teaching craft.

Context-Inappropriate Capability

High

Confidence: 97% confidence
Finding: The instruction to use sockpuppet-style comments such as '小号发' to simulate audience reaction is a deceptive engagement tactic intended to manipulate social proof and drive clicks. This is dangerous because it normalizes inauthentic behavior, can trigger platform enforcement, and can be scaled into coordinated manipulation or fraud-like promotion.

Context-Inappropriate Capability

High

Confidence: 98% confidence
Finding: The file explicitly teaches 'high imitation' of viral works by preserving plot skeletons and emotional beats while changing superficial details to avoid detection. In a writing skill, this is especially risky because it facilitates plagiarism evasion, copyright abuse, and mass production of derivative content disguised as original work.

Description-Behavior Mismatch

Medium

Confidence: 90% confidence
Finding: This chapter extends beyond fiction-writing instruction into operational cross-border monetization and tax/compliance guidance, which is outside the declared skill scope. That scope drift is risky because users may rely on unvetted financial and compliance instructions as if they were part of a safe writing curriculum, creating legal, tax, or platform-account consequences.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The file gives prescriptive instructions for completing a U.S. tax form, including specific treaty and rate selections, despite not being a tax-advice skill. This is dangerous because tax forms are jurisdiction- and fact-specific; incorrect filing could lead to false statements, withholding errors, regulatory issues, or tax liability for users who follow the template blindly.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The skill description is broad enough to match common user requests like learning to write, getting help generating text, or polishing content, which can cause unintended activation outside the narrow Fanqie-fiction context. This is dangerous because an over-broad skill may intercept unrelated conversations and steer them into its predefined workflow, reducing user control and potentially causing prompt-routing or instruction-precedence issues.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The quick-start mappings rely on highly ambiguous trigger phrases such as '想学习', '帮我生成', and '润色', with no scope constraints or negative examples. That makes accidental activation likely during ordinary conversations, allowing the skill to take over requests that are unrelated to Fanqie fiction and increasing the risk of incorrect routing and unintended behavior.

Natural-Language Policy Violations

Medium

Confidence: 93% confidence
Finding: The checklist presents Westernized names, titles, and setting elements as mandatory for overseas publishing, without framing them as optional market-specific adaptation or obtaining user opt-in. This can pressure users toward culturally flattening or identity-altering outputs and may lead the skill to override user preferences or produce exclusionary content when assisting with internationalization.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.