Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed AI-industry infographic workflow that mainly guides web research, fact-checking, prompt writing, and content packaging, with no hidden code or persistence.
Install this if you want a strict AI-industry infographic research and prompt workflow. Before using it, be aware that it defaults to Chinese/Xiaohongshu-style deliverables, performs many web searches, and may create or update local project files for generated packages and corrections.
The trigger list includes broad phrases such as “生成Prompt” and “做个选题的生图包,” which can match common user requests outside the narrow intended use case. That can cause the skill to activate unexpectedly and steer conversations into its rigid workflow, increasing the chance of undesired tool use, unnecessary web research, or off-scope content generation.
The skill description hard-codes Chinese-language deliverables and Chinese-market output conventions without checking the user's language or locale preference. This can override user intent and create misleading or unusable outputs, especially in multilingual environments or when the user expects neutral formatting.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
You MUST print the following report in your conversation response BEFORE the first Bash call that writes a file. This is the user's only way to verify rounds were completed without asking "did you run three rounds?"
══════════ 三轮核实报告 ══════════
The mandated four-section output format requires Chinese titles, Chinese social copy, and locale-specific formatting with no alternative path. This is dangerous because it forces a fixed communication channel and output language, which can cause policy or business-process errors when the user needs another language or a different platform format.
The section defines mandatory tone rules specifically for 小红书 content, including Chinese discourse markers and title formulas, as if they always apply. Without explicit user opt-in or region-specific justification, this creates a locale policy issue by forcing one platform/language style.
No suspicious patterns detected.