Security audit

本地向量记忆免费版

Security checks across malware telemetry and agentic risk

Overview

This local memory skill is mostly purpose-aligned, but it should be reviewed because it requests shell execution despite saying it is pure Markdown and it persists conversation details without clear retention or deletion controls.

Install only if you want the agent to maintain persistent local memory files for the workspace. Do not grant shell execution for this skill unless the publisher explains a real need for it. Review the created memory files regularly, avoid storing secrets or sensitive personal data, and delete SESSION-STATE.md, MEMORY.md, and memory/ entries when you no longer want them retained.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The manifest exposes the exec tool while the documentation claims the skill is pure Markdown and does not require command-line capability. This mismatch can mislead users and host agents into granting broader permissions than expected, increasing the attack surface if later instructions or prompt injection cause shell execution.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill repeatedly instructs the agent to persist user details, decisions, and preferences to local files across sessions, but it does not provide a clear warning or consent mechanism for data retention. This creates privacy and compliance risk because sensitive information may be stored indefinitely without the user's informed approval.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The skill directs the agent to retain natural-language user context and preferences in local memory files for reuse across sessions. Even without external exfiltration, this creates a real confidentiality risk because later prompts, other local users, backups, or unrelated tools may access and disclose stored sensitive data.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The workflow requires writing user-supplied details before replying and later migrating important content into long-term logs. This normalizes broad capture of conversation content and increases the likelihood that secrets, personal data, or confidential project details are preserved beyond the immediate task context.

Ssd 3

Low

Confidence: 86% confidence
Finding: The example operationalizes cross-session persistence of a user's preference and demonstrates automatic reuse later. While the stored example is low sensitivity, it reinforces a retention pattern that can easily be applied to more sensitive information without adequate safeguards.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.