Security audit

记忆雷达

Security checks across malware telemetry and agentic risk

Overview

This skill is a memory-file security scanner, but its broad activation language, unclear scan scope, and missing referenced scripts make its read/execute/write authority require review before installation.

Review this skill before installing. Use it only for explicit memory-security scanning, verify which files it will scan, confirm the referenced scripts actually exist, avoid enabling --allow-remote unless sending redacted memory context to an LLM is acceptable, and approve quarantine or cron setup only when you understand the exact files affected.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The skill documentation gives conflicting statements about what files are scanned: earlier sections claim coverage of memory files, logs, and workspace configuration, while the safety note limits scope to memory files only. This can cause operators to wrongly assume broader protection than actually exists, leading to missed threats in unscanned locations.

Intent-Code Divergence

Medium

Confidence: 88% confidence
Finding: The document contradicts itself on quarantine behavior: one section says threats trigger backup and line replacement after confirmation, while another says quarantine is never automatic. Ambiguity around destructive or modifying actions is dangerous in a security tool because users may authorize execution under false assumptions about whether files will be altered.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The invocation language is overly broad, covering generic safety, compliance, quality, and encryption-related development scenarios rather than a narrowly scoped memory-security task. In agent environments, broad trigger conditions can cause the skill to activate unexpectedly and run read/exec/write workflows on unrelated requests.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The example trigger phrase ('scan memory files for security issues') is generic enough to collide with normal user requests, especially in an agent platform where examples often influence routing behavior. This increases the chance of unintended skill invocation and execution-capable actions in contexts where the user did not intend to use this specific tool.

Natural-Language Policy Violations

Medium

Confidence: 80% confidence
Finding: The description frames Chinese interaction as a fixed behavior without offering language choice. While not directly enabling code execution, forced language constraints can cause misunderstanding of security prompts, scan results, or confirmation requests, increasing the chance of operator error.

Natural-Language Policy Violations

Medium

Confidence: 79% confidence
Finding: Presenting Chinese interaction as a core fixed capability without opt-in can reduce usability and clarity for non-Chinese users during security-sensitive workflows. Misinterpretation in warnings, quarantine confirmations, or restoration steps can lead to incorrect operator actions.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.