Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Seedance storyboard skill, but it includes explicit guidance to avoid copyright-related refusals, so it should be reviewed before use.
Before installing, review the Chinese-language templates and Seedance-specific output conventions. Use the skill only with original, licensed, or clearly permitted source material, and ignore or remove the instruction that suggests avoiding copyright refusals by renaming characters.
The file presents required production parameters using Chinese labels and examples, such as 写实/动画/水墨 and 16:9横屏 / 9:16竖屏, with no indication that the user can choose another language. This creates a locale/language constraint in the skill's instructions without explicit opt-in or justification.
The skill mandates Chinese field labels and filenames such as 图片1, 剧本.md, 素材清单.md, and 分镜.md, which effectively forces a specific language/locale in outputs. The policy exception does not apply because the file does not clearly justify the locale restriction or offer user opt-in.
This markdown file defines the role, workflow, templates, and output requirements entirely in Chinese, and it instructs the assistant to interact with users in that language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless a justified locale restriction is documented.
The instruction to use generic labels like 'Figure 1' and 'Figure 2' specifically to avoid copyright-related refusals is an evasion technique aimed at bypassing downstream safety or policy controls. Even though it does not directly execute code, it encourages users to conceal prohibited intent and increases the chance of generating infringing or disallowed content through another model.
The skill is entirely written in Chinese and provides no language selection, fallback, or explicit locale constraint. This can exclude or confuse users and downstream agents that expect multilingual or default-language compatibility, leading to misuse, misunderstanding of instructions, or failure to safely apply the tool in broader contexts.
No suspicious patterns detected.
