Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese-language advisory skill for government fund applications, with no executable code or hidden system-level behavior found.
Installers should treat this as a government-funding research aid, not legal, financial, or compliance advice. Before using its output in real submissions, verify current local rules with the relevant fund manager or government department and avoid overstating technical qualifications.
The skill metadata includes a long trigger list with several broad, high-frequency phrases such as government subsidy, project application, and application conditions. In an agent ecosystem, overly broad activation terms can cause the skill to trigger for loosely related conversations, leading the model to inject domain-specific guidance when the user did not intend to invoke this workflow.
The instruction suggesting projects without real technical barriers can be 'packaged' or reframed as 'AI+industry' encourages misrepresentation to satisfy government-fund eligibility criteria. In this skill context, that is dangerous because the agent is specifically designed to guide real funding applications, so users may be nudged into deceptive statements in regulated submissions, creating fraud, compliance, and reputational risk.
This markdown file is entirely written in Chinese and does not indicate that other languages are available or that Chinese is required for a region-specific reason. Under the language/locale policy rule, forcing a specific language without user opt-in can be a natural-language policy violation.
The activation guidance says to enable the skill when users ask broad questions like whether a project can get government money or which level to apply to, while also referencing a related funding-diagnosis skill. Because the boundary between this skill and adjacent funding-advice contexts is not crisply defined, an orchestrator may select it in situations better handled by another skill, producing misrouted or incomplete advice.
This markdown file contains user-facing guidance exclusively in Chinese, including headings, instructions, and decision criteria. Under the policy rule for natural-language violations, forcing a specific language without user opt-in can be a locale/language policy issue.
No suspicious patterns detected.
