Security audit

Memora Knowledge Base

Security checks across malware telemetry and agentic risk

Overview

This knowledge-base skill appears purpose-built, but its privacy claims do not clearly match its document upload, remote API, external AI, and web access behavior.

Review before installing. Use this only with a KB_API_BASE endpoint you trust, and treat uploaded files, search queries, created notes, and generated prompts as data that may leave your machine or reach cloud AI providers unless you have verified a fully local configuration. Avoid sensitive documents until the privacy mode, provider settings, backups, and deletion behavior are clear.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill declares only an environment requirement but the documentation clearly describes network access and API-driven behavior, including external LLM, embedding, and web-scraping functionality. Undeclared capabilities reduce transparency and can mislead users or policy engines about what the skill may access or where data may be sent.

Intent-Code Divergence

Medium

Confidence: 94% confidence
Finding: The documentation states the system is 'Fully Private' with 'no third-party uploads,' but elsewhere it documents use of DeepSeek, DashScope, OpenAI-compatible APIs, and optional rerank services. This is dangerous because users may upload sensitive documents under a false privacy assumption, causing unintended disclosure to external providers.

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: Claiming the system is 'Offline Ready' and works without internet conflicts with documented web scraping and dependence on external AI APIs for major features in common configurations. Users may rely on this claim in restricted environments and unintentionally trigger failing workflows or unexpected network egress.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill advertises web scraping and external LLM/API integration without prominent user-facing warnings about data transmission, target-site access, or privacy consequences. In a knowledge-base skill handling personal documents, this context increases risk because sensitive content may be sent to third parties or combined with scraped data without informed consent.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The documentation mentions permanent deletion but does not clearly emphasize that it is destructive and irreversible. In a knowledge management system, this can lead to accidental loss of documents, vector data, and related graph metadata, especially since deletion consistency across stores is a highlighted feature.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The upload path reads an arbitrary local file and transmits its full contents to whatever service is configured in KB_API_BASE, but the script provides no explicit privacy/security warning at the point of action. In a skill context, this increases the risk of accidental disclosure of sensitive local documents, especially if the endpoint has been reconfigured or is remote rather than localhost.

Missing User Warnings

Medium

Confidence: 79% confidence
Finding: Search, question-answering, and document-creation features intentionally send user-supplied text to the knowledge-base API, but the script does not clearly disclose that text leaves the local process and is transmitted to the configured service. This is primarily a transparency/privacy weakness rather than a classic exploit, but it can still matter when users submit sensitive notes or prompts.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.