ClawHub
Back to skill

Security audit

Deepseek Assistant Free

Security checks across malware telemetry and agentic risk

Overview

This is mostly a DeepSeek API helper, but it asks for broad agent capabilities and includes ambiguous callback/export language without clear data-handling boundaries.

Install only if you are comfortable with your agent using a DeepSeek API key and sending selected prompts or code to DeepSeek. Avoid using it with secrets, private files, regulated data, or broad project context unless you explicitly choose what is sent. Review the callback_url/export language and the unnecessary write permission before trusting automatic use.

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 (6)

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The skill describes callback_url support and 'create/query/export' style operations even though the advertised purpose is a simple DeepSeek API guidance/chat helper. This scope expansion can mislead an agent into handling asynchronous data delivery or broader data operations without clear trust boundaries, increasing the chance of unintended data exfiltration or misuse.

Intent-Code Divergence

Medium
Confidence
83% confidence
Finding
The dependency section conflicts with the rest of the skill by claiming the LLM API is platform-provided while examples require direct calls to DeepSeek with a user-supplied API key. This inconsistency can cause users or agents to misunderstand where data is processed and which credentials are being used, leading to accidental external transmission or incorrect trust assumptions.

Vague Triggers

Medium
Confidence
76% confidence
Finding
The invocation guidance uses broad trigger phrases such as general AI/agent/LLM use cases without tight scope limits. In agent environments, overbroad routing can cause the skill to be selected in contexts involving sensitive data, resulting in unnecessary execution of write/exec-capable instructions and possible off-system transmission.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The markdown documents callback URLs and external API request patterns without a clear warning that user prompts, code, or other content may be transmitted to third-party services. In this context, the skill also exposes exec capability, so the lack of disclosure meaningfully increases privacy and data-handling risk.

External Transmission

Medium
Category
Data Exfiltration
Content
const apiKey = process.env.DEEPSEEK_API_KEY;
// ...
async function chat(message) {
  const resp = await fetch('https://api.deepseek.com/v1/chat/completions', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${apiKey}`,
Confidence
90% confidence
Finding
fetch('https://api.deepseek.com/v1/chat/completions', { method: 'POST'

External Transmission

Medium
Category
Data Exfiltration
Content
const apiKey = process.env.DEEPSEEK_API_KEY;
// ...
async function chat(message) {
  const resp = await fetch('https://api.deepseek.com/v1/chat/completions', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${apiKey}`,
Confidence
90% confidence
Finding
https://api.deepseek.com/

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.