Deepseek

Security checks across malware telemetry and agentic risk

Overview

This looks like a legitimate DeepSeek/Membrane integration, but it gives agents broad authenticated API access without clear limits or confirmation rules.

Install only if you trust Membrane and need authenticated DeepSeek access. Prefer the named Membrane actions, avoid the raw proxy unless necessary, and require explicit user approval before sending sensitive prompts, files, billing/account data, or any write/delete request.

SkillSpector

By NVIDIA

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 (4)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The manifest advertises 'Manage Organizations' and broad interaction with DeepSeek data, but the body primarily enables model/completion use and generic API access. This capability mismatch can cause agents or users to invoke the skill under false assumptions, leading to over-privileged or unintended external operations beyond the declared purpose.

Context-Inappropriate Capability

Medium

Confidence: 97% confidence
Finding: The skill explicitly documents arbitrary proxy requests to DeepSeek endpoints, which creates a broad, effectively open-ended external action surface not justified by the stated purpose. In an agent setting, this can bypass the safer, enumerated action model and enable exfiltration, unintended data transmission, or use of undocumented/high-risk endpoints.

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The activation guidance is broad enough that an agent may invoke this skill whenever a request loosely relates to DeepSeek, even if the user's intent does not require external API interaction. Over-broad triggering increases the chance of unnecessary network access, data sharing, or use of a capability set the user did not mean to invoke.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The proxy-request section instructs direct API calls without requiring a user-facing warning or confirmation that data will be transmitted to an external service. In agent workflows, this can cause sensitive prompts, files, or metadata to be sent off-platform without sufficiently informed consent.

VirusTotal

48/48 vendors flagged this skill as clean.

View on VirusTotal