Security audit
Security checks for vulnerabilities and agentic risk
This appears to be a normal API-backed video generation skill, with privacy and quota considerations but no evidence of hidden or destructive behavior.
Install only if you are comfortable sending prompts and any referenced media to the video provider and using that provider's API key or credits. Review generated commands before execution, avoid sensitive prompts or private media URLs, and choose an explicit download location.
The skill advertises executable behavior that reads environment/config data, accesses local files, and performs outbound network calls, but it does not declare permissions or present these capabilities in a structured, least-privilege way. In an agent setting, hidden capability scope can lead to unintended secret access or file/network operations that users and policy layers cannot properly gate.
The documented purpose only partially reflects the actual operational behavior: the skill submits prompts to an external service, polls task status, downloads generated media, and retrieves API keys from multiple locations. This mismatch prevents informed consent and can expose user prompts, local secrets, and filesystem state to unexpected handling, especially when the prompt-authoring functionality is presented more prominently than the external execution path.
The skill can be triggered by very broad, generic phrases like '视频脚本' or '生成视频', which increases the chance of accidental or adversarial invocation outside the user's intended context. Because the skill can move from prompt-writing into direct API-based video generation, an overbroad trigger surface may cause unintended external actions, quota consumption, or processing of unreviewed user input.
The trigger phrases are broad enough to match ordinary conversation about videos, scripts, or prompt writing, increasing the chance that the skill activates in contexts where the user did not intend tool-assisted generation or API-backed actions. Over-broad invocation raises the risk of accidental escalation from advice to networked execution and file creation.
The markdown instructs users to directly invoke an API-backed script that transmits prompts and optional image URLs to a third-party service and writes resulting media to local storage, but it does not foreground privacy, cost, or file-output warnings. This makes accidental disclosure of sensitive prompts or references more likely and obscures that generated artifacts will be downloaded and stored automatically.
