Security audit

Win Dev Pitfalls Publish V1.72.0

Security checks for vulnerabilities and agentic risk

Overview

This markdown-only skill is mostly a Windows troubleshooting vault, but it needs Review because it includes detailed anti-termination, watchdog, privileged repair, and kernel-driver self-protection guidance beyond its broad description.

Install only if you intentionally want an agent to work on Windows defensive-security tooling and are comfortable reviewing self-protection, privileged repair, service, driver, and system-policy guidance before use. Treat any hosts, Defender, registry, firewall, UAC, service, watchdog, TESTSIGNING, or kernel-driver action as requiring explicit user approval and preferably a VM or lab environment. If you only need ordinary bat/PowerShell/Flutter troubleshooting, remove or split out the self-protection and security-tool sections first.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (4)

Intent-Code Divergence

High

Confidence: 97% confidence
Finding: The document claims it only stores de-identified technical experience, but substantial later content covers anti-termination, watchdog, mitigation bypass, and self-protection techniques that materially expand from neutral troubleshooting into dual-use evasion guidance. This mismatch is dangerous because it lowers reviewer suspicion and can normalize or conceal offensive content inside a seemingly benign 'pitfalls' knowledge base.

Context-Inappropriate Capability

Critical

Confidence: 99% confidence
Finding: These sections document malware-adjacent self-protection capabilities such as watchdog restart logic, anti-termination hardening, DLL-loading failure handling under hostile environments, and process mitigation tuning. In a generic Windows pitfalls skill, this context makes the content more dangerous because it provides operational guidance that can directly improve resilience and stealth of unwanted software, not just benign debugging.

Context-Inappropriate Capability

High

Confidence: 95% confidence
Finding: The skill includes actionable instructions to modify hosts files, trigger elevation, alter system security behavior, and use encoded PowerShell command delivery. Even when framed as remediation, this goes beyond neutral troubleshooting and creates reusable patterns for privileged system tampering that can be repurposed offensively.

Context-Inappropriate Capability

Critical

Confidence: 100% confidence
Finding: This section discusses kernel-driver-assisted self-protection and blue-screen-based anti-termination behavior, which is highly sensitive dual-use material with little justification in a general development pitfalls vault. Such guidance can be directly abused to make malicious software harder to stop, recover, or analyze, significantly raising harm potential.

Static analysis

No suspicious patterns detected.