Security audit
Security checks for vulnerabilities and agentic risk
This Markdown-only Seedance prompt helper is mostly coherent, but it includes explicit copyright/platform-review avoidance guidance that users should review before installing.
Use this only if you want a Chinese-focused Seedance prompt-writing helper. Before installing, review or remove the IP/copyright avoidance guidance and avoid using it to imitate protected brands, characters, or styles; use original or properly licensed references instead. Also be aware that the broad trigger terms may affect ordinary video, storyboard, or product-promo requests.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
Mode: All-Reference
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
Mode: All-Reference
The manifest description includes broad trigger terms like video generation, prompt, and product showcase that can match many ordinary requests and cause the skill to activate outside its intended niche. Over-broad activation can unexpectedly steer unrelated conversations into this skill's rigid prompting workflow, degrading user intent handling and creating prompt-scope hijacking risk.
Declaring the skill as Chinese-optimized without offering a language choice can force behavior that does not match user preference and may reduce transparency or usability for non-Chinese users. This is primarily a policy and UX scoping issue rather than a direct security exploit, but it can contribute to inappropriate skill activation and user confusion.
The read_when list uses short, generic phrases such as 分镜设计, 镜头语言, 电商视频, and 产品展示视频 without qualifiers, which are common across many benign creative tasks unrelated to this skill. This makes unintended invocation likely and can cause the agent to apply specialized constraints and outputs when the user did not request Seedance-specific assistance.
The instruction that users describe needs in Chinese imposes a language constraint without consent, which can override user intent and create mismatched outputs. In context this is not highly dangerous, but it is still an undesirable hard-coded behavior that may cause the agent to mishandle multilingual or English-language requests.
The IP/版权规避 section gives stepwise instructions for progressively evading platform moderation while preserving recognizable copyrighted aesthetics, including what to do after rejection. That is dangerous because it operationalizes policy evasion and enables users to generate infringing or disallowed derivative content despite platform safeguards.
The phrase 'IP规避' explicitly signals functionality aimed at avoiding intellectual-property restrictions, which can enable copyright evasion, trademark mimicry, or deliberate bypass of safety and policy controls around protected content. In the context of an AI video prompt skill, this makes the skill more dangerous because it operationalizes generation workflows that may help users produce infringing or policy-violating outputs at scale.
The description states the skill is a '中文优化版' (Chinese-optimized version), which imposes a language/locale constraint in the natural-language metadata. There is no indication that users can choose another language or opt into this locale behavior, which matches the language-policy concern for this rule.
Line L003 says to use the following English terms in Seedance 2.0 prompts, which imposes a specific language requirement. The file does not indicate that other languages are acceptable, nor does it provide a user opt-in or a documented reason for the constraint.
SQP-3 applies to all file types and covers language or locale policy violations. Line L003 presents the template library as directly usable with Chinese-only explanatory annotations, with no indication of alternative language support or user opt-in.
No suspicious patterns detected.