Security audit

Dlazy Generate

Security checks for vulnerabilities and agentic risk

Overview

This skill is mostly a disclosed media-generation CLI wrapper, but it needs Review because it can run external commands, upload local media to a hosted service, and save an API key while its activation and setup instructions are under-scoped or inconsistent.

Before installing, confirm you want an agent to run the dLazy CLI for media generation and send prompts or selected local media files to dLazy-hosted services. Use only media and voices you have rights or consent to process, prefer a per-invocation DLAZY_API_KEY if you do not want persistent key storage, and verify the intended @dlazy/cli version before installing globally.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill gives contradictory authentication guidance: one section says all requests require a dLazy API key, while a later section says no additional API key is needed except explicitly marked external APIs. In an agent-executed skill, inconsistent auth semantics can cause unsafe execution attempts, accidental credential prompts, or user confusion about where secrets are required and should be stored.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 86% confidence
Finding: The metadata says the skill is not suitable for copyrighted media processing, but the documented capabilities include dubbing, lip-sync, and media replication of existing assets. That mismatch can mislead users and agents into believing the skill is safer or more restricted than it really is, increasing the chance of policy-violating or unauthorized media handling.
Content: No source excerpt is available for this finding.

Vague Triggers

High

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The trigger keywords include broad everyday terms like 'generate' and 'multimodal generation', making accidental invocation likely in normal conversation. In a skill with exec capability, overbroad activation increases the chance that shell commands are run without the user intending to invoke this specific toolchain.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 94% confidence
Finding: The skill instructs the agent to run shell commands (dlazy <model_name> -h and then execute the model command) without a prominent upfront warning that this will invoke external binaries, transmit prompts/media to remote services, and may upload local files. This creates a meaningful risk of unintended command execution and data exfiltration in an agent environment.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.