Security audit
Security checks for vulnerabilities and agentic risk
This skill is a Chinese-first prompt-generation helper, with some broad activation wording but no hidden access, persistence, networking, or destructive behavior found.
Install this if you want a Chinese-first assistant for AI art and video prompts. Be aware it may activate for broad image/video-generation wording and often emits Chinese plus English fields, so users who need strict language preference control may want clearer locale guidance from the publisher.
The file presents the skill documentation entirely in Chinese, beginning with the title "更新日志", with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without opt-in can be a natural-language policy concern unless the locale restriction is explicitly justified.
The changelog explicitly documents adding broad trigger phrases and even emphasizes that the skill 'must be used' for common prompt-related requests. In an agent skill system, overly broad activation criteria can cause inappropriate routing, override user intent, and increase the chance that the skill intercepts unrelated requests.
The trigger expansion section adds generic terms such as 'AI绘画', '图像生成', '视频生成', and '优化提示词' without boundaries or disambiguation. In a multi-skill environment this can create routing collisions and prompt hijacking behavior, where this skill captures broad categories of normal user requests that may belong to other tools or require direct assistant handling.
The README is entirely written in Chinese and presents output expectations such as Chinese-labeled fields and Chinese video prompt guidance, but it does not state that users may choose another language. Under the policy, forcing a specific language without opt-in is a natural-language policy concern unless clearly justified or optional.
The trigger guidance is broad enough to activate the skill for generic image/video discussions, not just explicit prompt-generation requests. In an agent system, this can cause unintended routing, override a more appropriate skill, or steer users into generation workflows they did not ask for, which is a real security/reliability issue even without malicious intent.
The skill metadata and description state that when users mention broad phrases like “生成提示词”, “优化提示词”, “AI绘画”, “图像生成”, or “视频生成”, this skill must be used. These triggers are expansive and can cause unintended activation on loosely related requests, increasing the chance that the assistant routes user input into rigid skill behavior that overrides better contextual handling or safer alternatives.
The required JSON schema forces both Chinese and English fields for generated output, regardless of whether the user asked for bilingual content. This can violate user expectation, cause unintended language transformation, and in some contexts leak or restate sensitive user-provided content into another language without explicit opt-in.
This JSON manifest contains multiple natural-language prompt and output requirements exclusively in Chinese, including explicit expectations like translating into Chinese and avoiding Chinese confirmation words. For a general-purpose skill/eval file, this creates a language policy constraint without offering user choice or documenting a justified region-specific scope.
The entire example, including the title, input, and output, is presented only in Chinese and does not offer any language choice or indicate that the skill is intentionally limited to Chinese users. This creates a natural-language locale policy concern because it implicitly enforces a specific language without user opt-in.
The file’s title and example content are entirely in Chinese, which establishes a single-language behavior without offering the user any language or locale choice. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is clearly documented and justified.
The en value on L3 begins with and substantially includes Chinese text rather than providing an English-localized prompt. This can violate language/locale expectations because the file distinguishes cn and en fields but does not actually give users an English-language option in the en entry.
The usage and example output messaging are presented only in Chinese, and the file name and docstrings also indicate a Chinese-first experience. Under the policy rule, forcing a specific language without user opt-in can be a natural-language policy violation when no alternative or choice is offered.
