Secure Communicator

• Outcome: Review
• Latest audit: May 29, 2026, 9:34 PM
• Version:
  1.0.1

Overview

This skill is a local encryption helper, but it makes strong security claims while using weak custom cryptography and incomplete safeguards for sensitive data.

Treat this as a Review item rather than malware. It does not show exfiltration or hidden execution, but you should not rely on it for real confidential communication or high-value secrets. Prefer vetted tools such as age, OpenPGP, or libsodium-based encryption, and only install this if you understand the custom-crypto limitations and can manage output files and key files securely.

SkillSpector

By NVIDIA

Vulnerability Patterns

• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Intent-Code Divergence

High

• Confidence: 99% confidence
• Finding: This is a real vulnerability. The `encrypt`/`decrypt` path generates per-message keys, places both `payloadKey` and `metadataKey` directly into the serialized output, and does not use the caller-supplied key file for protection, so anyone with the ciphertext can recover the plaintext. In a skill explicitly marketed for securing sensitive messages over insecure channels, this is especially dangerous because users are likely to trust it for confidentiality when it provides effectively none.

Intent-Code Divergence

Medium

• Confidence: 82% confidence
• Finding: The comments and surrounding documentation are misleading relative to the actual behavior, including claims of a 'correct implementation' and metadata semantics that do not fully align with what is written. Misrepresenting cryptographic behavior is security-relevant because maintainers and users may rely on incorrect assumptions, increasing the chance that the insecure API is used or that protections are overestimated.

Natural-Language Policy Violations

High

• Confidence: 98% confidence
• Finding: The skill markets a homegrown 'triple-layer XOR' scheme as suitable for sensitive communications over insecure channels, which can mislead users into relying on weak, nonstandard cryptography for real confidentiality. Repeated-key XOR with custom packaging is not comparable to vetted modern encryption, and the surrounding language ('secure', 'end-to-end encryption', use over email/chat/Telegram) materially increases the chance of unsafe deployment.

Missing User Warnings

Medium

• Confidence: 81% confidence
• Finding: The file encryption/decryption examples instruct users to write outputs to paths like 'encoded.txt' and 'decrypted.pdf' without warning about overwriting existing files, writing sensitive plaintext to insecure locations, or accidentally exposing decrypted data. In a skill explicitly handling sensitive messages and files, omission of these safeguards increases the risk of data loss or confidentiality breaches through operator error.

VirusTotal

63/63 vendors flagged this skill as clean.