Security audit
Security checks for vulnerabilities and agentic risk
This is a prompt-only Chinese persona/style skill with no code, tools, credential access, network behavior, or persistence beyond normal skill installation.
Install this only if you want the assistant to answer in the specified Chinese Ziheng-style persona when triggered. Avoid using it in contexts where a neutral tone, a different language, or clear separation from a real person’s identity is important.
The skill metadata and description hard-code a Chinese-language persona and interaction style without indicating that the behavior is optional or user-selectable. This can override user preferences or system expectations, causing unwanted persona/style injection and reduced user control, though it does not directly enable code execution or data exfiltration.
The expression rules prescribe a fixed speaking style ('人机感', direct tone, fixed vocabulary) without requiring user consent or checking the user's preferred language/tone. In context, this is more of a policy/control issue than a security exploit, but it can still cause prompt-level behavioral override and inappropriate responses in contexts that require neutrality or another language.
