Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent daily-report helper, but it asks the agent to remember sensitive workplace context without clear consent, limits, or deletion controls.
Review this skill before installing if your daily reports include confidential project names, customer details, internal incidents, financial metrics, or personnel information. Use it only with information you are comfortable having reused for personalization unless the publisher adds explicit opt-in memory, retention limits, and deletion controls.
The skill description is written entirely in Chinese and frames the product for Chinese-speaking office workers, with no indication that users may choose another language or locale. The policy requires flagging language or locale constraints when a skill forces a specific language without user opt-in.
The skill description and all examples are written as Chinese-only triggers and workflows, and the skill does not offer any language or locale choice. This can violate language/locale policy when a skill implicitly forces a specific language without user opt-in or a documented regional justification.
The skill explicitly says it will remember the user's job background, project names, terminology, and prior report style, but provides no warning, consent flow, or limits on retention. Because daily reports often contain sensitive workplace details, persistent reuse across sessions can expose confidential employer or personal information beyond the user's immediate request.
This instruction creates a natural-language data retention mechanism for workplace background, project names, and prior report preferences. Such retained context can include internal company information, and if surfaced later to the wrong conversation or used without user awareness, it creates a real privacy and confidentiality risk.
The manifest explicitly states that weekly and monthly reports are not applicable and should use a different skill, but the file defines a '周汇总模板（周报辅助）' with concrete weekly-report output structure. That extends the skill beyond its declared日报-only scope and creates a semantic mismatch between advertised purpose and actual documented behavior.
The 'Learning & Memory' section states the assistant will remember and learn from user work context, preferred styles, project names, and prior report highlights without disclosing storage or reuse implications. In a workplace-reporting skill, this can accumulate sensitive operational details and increase the chance of cross-session leakage or unintended disclosure.
The ongoing 'remember, learn, improve' model encourages accumulation of user and employer context across interactions, including prior report highlights and common terminology. In this skill context, that materially raises the risk of storing sensitive business information and later disclosing it inappropriately or beyond the user's expectations.
The manifest says weekly reports are not supported and should use another skill, but the version record says a future version plans '+周报自动汇总'. This documentation points in the opposite direction of the declared scope and signals intent that contradicts the skill's stated boundaries.
