meetmind

• Outcome: Review
• Latest audit: Jun 1, 2026 · 5:41 PM UTC
• Version: 1.0.1

Overview

MeetMind appears to be a real meeting-notes client, but it sends sensitive meeting content and persistent identifiers to a cloud service without enough privacy and control detail.

Install only if you are comfortable sending meeting text, audio, user identifiers, and possibly long-lived meeting memory to hermesai.ltd. Avoid using it for confidential, regulated, or client-sensitive meetings unless the publisher provides acceptable privacy, deletion, retention, and certificate-handling terms.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (7)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill declares no permissions while documenting capabilities that access environment variables, local files, networking, and likely shell-assisted operations. This undermines informed consent and review because users and platforms cannot accurately assess what data and system resources the skill can touch before installation or execution.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 94% confidence
• Finding: The documented purpose focuses on meeting summarization, but the file describes broader behavior: persistent user ID generation, identity/header forwarding, account and usage queries, certificate activation, enterprise template CRUD, and possibly local transcription. This mismatch is dangerous because users may provide sensitive meeting content without realizing the extent of identity handling, storage, remote account operations, and local processing performed by the skill.

Context-Inappropriate Capability

Medium

• Confidence: 91% confidence
• Finding: The code pulls ambient identity from environment variables and persists a stable local identifier to disk, creating cross-session tracking and potential identity confusion without explicit user consent. In a meeting assistant context, this is sensitive because meeting notes, transcription jobs, and memory retrieval are tied to that identifier, so misbinding or silent persistence can expose another user's usage data or historical context.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill explicitly sends meeting text or audio and user identity to cloud endpoints, but the documentation does not provide a clear privacy notice covering what is transmitted, how long it is stored, who can access it, or whether it is used for model training or analytics. Because meeting content commonly contains confidential business and personal data, this omission creates material privacy and compliance risk.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill documents automatic analytics events containing identifiers such as skill_id and user_id, but does not present a user-facing warning, consent flow, or opt-out. Silent telemetry on a tool handling sensitive meeting data increases privacy exposure and can create regulatory issues, especially when tied to persistent user identity.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The function uploads local audio files to a remote service for transcription, but the operation site does not present any explicit notice, consent gate, or data-handling warning. Because meeting audio commonly contains confidential business discussions and personal data, silent cloud upload materially increases privacy and compliance risk.

Missing User Warnings

Low

• Confidence: 84% confidence
• Finding: The code writes a persistent user identifier to a local dotfile without informing the user, enabling silent tracking across runs. While the stored value is not a secret by itself, it is still identifying metadata that can link meeting activity and cloud-side history over time.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.