Security audit
Security checks for vulnerabilities and agentic risk
This tarot skill is content-only, but it should be reviewed because it steers users toward private messages and more personal details without privacy safeguards.
Review this skill before installing if you run tarot or social-media reply workflows. It has no code execution or system access, but it may encourage users to disclose relationship or emotional details in private messages; add clear privacy, data-minimization, and non-professional-advice boundaries before using it with real users.
The trigger list is broad enough to match general divination or card-reading requests, not just tarot-specific ones, which can cause the skill to activate outside its intended scope. Over-broad routing increases the chance of irrelevant or misleading responses and can interfere with safer or more appropriate skills selected for the user’s actual request.
The skill specifies Chinese as the primary language without checking the user’s language preference, which can lead to unexpected language switching and poor comprehension. In a multi-skill environment, this creates a usability and policy-alignment risk because the system may disregard the user’s current language context or accessibility needs.
This markdown file defines user utterances like “帮我抽一张” and “今天运势怎么样” as activation examples, but it does not specify whether these are exhaustive triggers, context-limited, or exclude ordinary conversational use. Because the phrases are common and broad, they could overlap with everyday chat and cause unintended invocation of the skill.
The skill explicitly invites users to continue in private messages and even requests increasingly specific personal details, but provides no warning about privacy implications or data handling. This is risky because users may disclose relationship, emotional, or other sensitive information under the impression they are receiving trusted guidance, with less public accountability once moved off the comment thread.
The private-message escalation language is open-ended and encourages moving users into a less visible channel without clear qualification criteria. In a divination or advice context, this can facilitate collection of sensitive personal details and reduce transparency or oversight around how user disclosures are handled.
This knowledge file provides authoritative-sounding tarot interpretations across sensitive domains like relationships, career, and finances without any warning that the content is entertainment or reflective guidance rather than reliable decision support. Users may treat the outputs as prescriptive advice and make consequential life decisions based on unverifiable spiritual claims, which is especially concerning because the skill is explicitly designed for divination use.
This markdown file presents all instructional/reference content exclusively in Chinese, and there is no indication that users can choose another language or that the skill is intentionally limited to a Chinese-language audience. Under the language/locale policy rule, forcing a specific language without user opt-in can be a natural-language policy violation.
