Dlazy Text Storyboard Script

• Outcome: Review
• Latest audit: Jul 21, 2026 · 4:03 AM UTC
• Version: 1.2.3

Overview

This skill says it is a text-only storyboard generator but also instructs agents to install and run a cloud media-generation CLI.

Install only if you intend to use dLazy cloud generation as well as text storyboarding. Review the npm package and API-key storage behavior first, avoid providing sensitive copy or private local media paths, and require explicit confirmation before any dlazy command is run.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (10)

Description-Behavior Mismatch

High

• Confidence: 98% confidence
• Finding: The skill explicitly presents itself as a text-only storyboard generator, but later includes instructions to execute terminal commands and generate images. This scope mismatch is dangerous because it can cause an agent or operator to grant tool access and perform networked media generation that users did not request or consent to, increasing the attack surface beyond the declared purpose.

Intent-Code Divergence

High

• Confidence: 97% confidence
• Finding: The file first states that no tools should be invoked to generate image/video/audio assets, then later mandates terminal-based generation steps. Contradictory instructions like this are risky because downstream agents may follow the more operationally specific section and execute external commands despite the earlier safety boundary, leading to unexpected tool use and data transfer to third-party services.

Context-Inappropriate Capability

High

• Confidence: 99% confidence
• Finding: The terminal execution workflow is not justified by the skill's stated purpose of generating storyboard scripts, yet it directs the agent to run model-generation commands and return hosted asset URLs. Unnecessary command execution and external API use are dangerous because they can trigger unreviewed network actions, upload local files, consume credentials, and create opportunities for prompt-to-action abuse.

Description-Behavior Mismatch

High

• Confidence: 98% confidence
• Finding: The skill presents itself as a storyboard-script generator and explicitly states it should not invoke tools to generate media, but later instructs the agent to collect image requirements and execute CLI commands to generate images. This mismatch can trick an orchestrator or user into granting broader execution and network/upload behavior than the declared scope, enabling unintended command execution and data transfer.

Intent-Code Divergence

High

• Confidence: 98% confidence
• Finding: The document explicitly says the skill is only responsible for script generation and must not generate assets, yet later includes 'CRITICAL EXECUTION INSTRUCTIONS' directing terminal-based image generation. Contradictory instructions of this kind are dangerous because they obscure the real behavior of the skill and can bypass human review or policy gating based on the declared function.

Context-Inappropriate Capability

Medium

• Confidence: 92% confidence
• Finding: A storyboard script generator does not need shell access to run iterative generation commands, yet the skill directs the agent to execute terminal commands against an external CLI. Unnecessary execution capability increases the attack surface and creates opportunities for command misuse, unauthorized network calls, or scope creep beyond the user's expected task.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The skill describes terminal execution and external API/media upload behavior, but the step-by-step task flow does not clearly foreground these operations as consent-sensitive actions before they occur. This can lead users to provide prompts or local file paths without understanding that data may be uploaded to third-party services or that commands will run locally.

Ssd 3

Medium

• Confidence: 90% confidence
• Finding: Requiring the model to retain and reproduce all user-provided text word for word removes any opportunity to minimize, redact, or safely transform sensitive content. If users include secrets, personal data, confidential drafts, or regulated information, the skill is instructed to echo it back in full, increasing disclosure risk.

Ssd 3

Medium

• Confidence: 90% confidence
• Finding: The requirement that shot text concatenate to exactly equal the original input mandates complete reproduction of all submitted content. That design prevents privacy-preserving handling and can expose sensitive material in generated outputs, logs, downstream tools, or shared storyboard documents.

Ssd 3

Medium

• Confidence: 91% confidence
• Finding: Directing the model to place the user's original text word-for-word into the 'Spoken Script' output creates a straightforward path for reproducing private or dangerous content verbatim. In contexts where outputs are stored, reviewed, or forwarded, this amplifies the risk of accidental leakage and policy violations.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.