Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed philosophy/thought-experiment prompt skill with broad auto-activation, but it contains no code, credential handling, data access, remote loading, or destructive behavior.
Install this only if you want a globally available philosophy/AI-ethics response style that may automatically influence related conversations. Users who prefer explicit opt-in behavior should narrow triggers, disable always_on, or invoke it only with /kh.
The trigger keywords include very broad terms such as 哲思, 悖论, 认知, 意识, and AI伦理, which overlap with ordinary conversation topics. In a skill system, this can cause unintended activation and prompt steering during unrelated user requests, reducing user control and increasing the chance of response hijacking by the skill.
The skill is configured as global, high-priority, and always_on, but its activation boundaries are not tightly constrained. This combination makes the skill persistently eligible to influence many interactions, which can override user intent, interfere with other skills, and create a standing prompt-injection surface across broad topic areas.
The prompt instructs the system to automatically link into broad categories like philosophy, consciousness, ethics, and AI whenever such topics arise. Because these are common and ambiguously defined domains, the skill can silently steer normal conversations and inject its own behavioral policy into requests that did not explicitly ask for it.
The skill claims it will automatically identify broad categories of topics such as philosophy, AI, ethics, and consciousness, which implies unsolicited activation across a wide range of normal conversations. This can cause unexpected behavior, override user intent, and expand the skill's influence beyond explicit invocation, especially because the trigger scope is vague and not narrowly bounded.
The README states the skill uses a "Unique cold & rational writing style," which imposes a specific communication style by default. This can conflict with language/locale or communication-policy expectations when the user is not given a choice or opt-in.
The prompt mandates a uniform “KAI world-view tone” and a fixed writing style for responses, without indicating that users can choose another language or locale style. This is a natural-language policy concern because it imposes a presentation constraint unilaterally rather than offering user preference or opt-in.
The documentation states the skill will maintain a fixed 'cold, speculative' persona for the entire interaction, without indicating user control or a way to switch styles. While not severe on its own, this can reduce user agency, make outputs less appropriate for context, and combine poorly with automatic activation by imposing behavior the user did not request.
No suspicious patterns detected.
