Security audit
Security checks for vulnerabilities and agentic risk
This skill is a Markdown-only Chinese relationship-chat advisor that relies on user-provided chat history and does not install code, access files, use the network, or persist outside the conversation.
Before installing or using this skill, avoid pasting raw private conversations unless needed. Redact names, phone numbers, account IDs, addresses, secrets, and intimate or third-party details where possible, and prefer summaries when exact logs are unnecessary.
The activation description is broad enough to match ordinary requests about chatting, relationship context, or reply advice without clear boundaries. This can cause unintended activation and collection of sensitive interpersonal data, especially since the skill is designed to ingest chat logs and maintain per-person profiles.
The skill description invites handling of intimate relationship background and chat history but does not warn users that these materials may contain sensitive personal or third-party data. That omission increases the risk that users will paste private conversations without understanding the privacy implications or minimizing exposed information.
The trigger examples like '帮我回' or '怎么回' are highly ambiguous and may overlap with many unrelated conversational tasks. In this skill, ambiguous triggers are more dangerous because they can lead to automatic context loading, profile creation, or sensitive chat-log processing for the wrong person or in the wrong context.
The instruction '所有输出始终使用中文' forces a specific language for all interactions. Under the policy, language restrictions should either be optional for the user or clearly justified as a region-specific requirement; neither is present here.
The onboarding explicitly asks the user to paste recent chat logs, but provides no caution about redacting private information, obtaining consent where appropriate, or avoiding disclosure of third-party sensitive data. Because this is the first-run flow, it directly encourages oversharing at the highest-risk moment before any safeguards are presented.
No suspicious patterns detected.
