Security audit

Dlazy Search Image

Security checks across malware telemetry and agentic risk

Overview

This skill is a dLazy cloud CLI wrapper whose documentation does not consistently match its advertised Pixabay image-search purpose.

Review this before installing. Use it only if you are comfortable installing a general dLazy CLI, storing or providing a dLazy API key, and sending search terms or referenced inputs to dLazy cloud services. Avoid passing local files or piped sensitive data unless you intend to upload them, and verify the correct argument name before automating the command.

SkillSpector

By NVIDIA

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Description-Behavior Mismatch

High

Confidence: 97% confidence
Finding: The skill metadata and opening description present this as a narrow Pixabay image-search tool, but the body states it sends prompts to dLazy APIs and may upload local media to dLazy storage. That mismatch can mislead an agent or user about what service is actually contacted and what data leaves the host, creating a trust-boundary violation and possible unintended data disclosure.

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The documented interface includes capabilities such as pipe-referenced inputs, local file upload, and saving remote results to an arbitrary local path, which exceed the stated purpose of simple image search. Hidden or under-declared I/O capabilities increase the attack surface because an agent may pass sensitive local content upstream or write files locally under the guise of a benign search operation.

Intent-Code Divergence

High

Confidence: 98% confidence
Finding: The documentation inconsistently claims the skill uses the Pixabay API while later stating it is a wrapper around dLazy-hosted APIs. Conflicting provenance and backend descriptions prevent informed consent and make it difficult to assess where queries, credentials, and files are sent, which is especially risky for agentic tools that may handle sensitive input.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill metadata and description present this as a Pixabay keyword image search tool, but the body discloses it is actually a thin client for dLazy-hosted APIs that may transmit prompts and local files to third-party endpoints. This mismatch can mislead users and agents about data flow, trust boundaries, and provider provenance, causing accidental disclosure of sensitive inputs to an unexpected service.

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The skill is advertised as a simple keyword-based image search tool, but the interface and workflow resemble a broader generation/inference client with async task handling and prompt-style semantics. That discrepancy increases the risk that users or agents will supply richer or more sensitive content than intended under the assumption they are only performing a bounded search operation.

Intent-Code Divergence

Medium

Confidence: 88% confidence
Finding: The documented command options define --query, but the examples and error text refer to --prompt as required. This inconsistency can cause invocation failures, user confusion, and unsafe agent retries or workarounds, especially in automated workflows that may substitute unintended fields or mishandle user data.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.