Security audit

综合生成工具-免费版

Security checks across malware telemetry and agentic risk

Overview

This skill is mainly an image-generation helper, but it asks agents to activate it for much broader AI and automation tasks than its stated purpose supports.

Review before installing. Use this only for dlazy image generation or image editing tasks, and avoid letting it handle general chat, agent orchestration, file search, or arbitrary automation. If installed, prefer an environment variable for the API key when possible, restrict reference images to intended files, and be aware that the global npm install and dlazy CLI will run third-party code and call an external service.

SkillSpector

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The manifest and description position this as a lightweight image-generation tool, but the skill text broadens scope to generic AI model calling, intelligent chat, agent orchestration, and LLM applications. This kind of scope inflation can cause inappropriate activation and over-trust, leading an agent to invoke a higher-privilege skill for tasks unrelated to image generation.

Description-Behavior Mismatch

Medium

Confidence: 92% confidence
Finding: The documented core functionality claims broad automation, file processing, API integration, command execution, and information retrieval, which materially exceed the declared purpose of an image tool. In an agent setting, these claims expand the apparent authority of the skill and can justify unsafe use of read/write/exec capabilities beyond the intended task domain.

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: Claiming generic command execution, file processing, and information retrieval in a skill that should mainly broker image-generation API calls creates dangerous ambiguity around what the agent may do locally. Because the skill is allowed to use exec/read/write, this mismatch increases the risk of the model performing unnecessary local actions or accessing unrelated data under the banner of 'core functionality.'

Vague Triggers

High

Confidence: 95% confidence
Finding: The trigger condition '需要AI模型调用、智能对话、Agent编排、LLM应用时使用' is so broad that it can match a large fraction of unrelated AI tasks. Overbroad activation is dangerous because it can route general-purpose requests into a skill with exec/read/write permissions and external API access, increasing the chance of data exposure or unauthorized actions.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The keyword-based scope description is extremely broad and ambiguous, using generic phrases that can match many unrelated prompts. In agent environments, vague scope keywords increase accidental or adversarial activation, especially when the skill also advertises local execution and file capabilities.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.