Security audit
Security checks for vulnerabilities and agentic risk
This script-writing skill requires an unrelated image-generation workflow that adds software execution, account credentials, and external data sharing.
For script writing alone, prefer a version that removes the mandatory image-generation workflow. If you also want dLazy images, explicitly approve that separate task and review software installation, prompt and media sharing, possible service charges, and local API-key storage. The concrete concern is the workflow mismatch; malicious behavior is not established.
The supplied command performs authentication setup, with no demonstrated connection to the declared script-generation purpose. Nothing in the chunk establishes this as a supporting implementation detail of that workflow. The code does not show credential exfiltration or actual script generation.
The supplied code contains only a login command. Authentication to an external tool is not described in the declared purpose, and nothing in the supplied chunk connects it to script generation. The code does not establish whether login succeeds or what resources the tool accesses.
The supplied command performs authentication configuration rather than script generation or writing guidance. No connection between dlazy authentication and the declared purpose is established. The snippet does not show script generation, credential exfiltration, or other behavior beyond the authentication command.
The supplied code contains only a login command, with no script-writing guidance or generation behavior. Authentication to dlazy is not described by the declared purpose or triggers, and no connection to the script-generation workflow is established. The snippet does not reveal the login mechanism or whether authentication succeeds.
The installation metadata introduces an external npm executable, while the final instructions mandate terminal-based image generation despite the skill being advertised as text-script guidance. This unnecessarily expands a text-only task into third-party code execution and network requests; a compromised package could act with the agent's operating-system permissions, although no malicious package behavior is demonstrated here.
This duplicates the external-execution scope mismatch: a task that can be completed entirely as text is redirected toward an npm CLI and hosted image-generation service. Version pinning and source links improve auditability but do not establish package safety or justify executing it for the requested task.
The skill's mandatory execution instructions redirect a text-only script-writing task into image generation and require terminal execution of a third-party CLI. This unnecessarily expands the task's authority and data exposure, potentially transmitting prompts to an external service and consuming paid API resources without a need arising from the original request. The disclosed service endpoints and explicit confirmation steps reduce the risk of covert execution, but do not resolve the unrelated workflow or its conflict with the requirement to directly output script text.
The final 'CRITICAL EXECUTION INSTRUCTIONS' contradict direct script output and require an unrelated image-generation workflow with terminal execution. Although image generation requires user confirmation, that confirmation does not necessarily authorize installing software or transmitting potentially sensitive prompt content to a third party.
No suspicious patterns detected.
