Security audit
Security checks for vulnerabilities and agentic risk
This is a small Chinese-language Mao Zedong Thought Q&A skill with no network, credential, persistence, or system-modifying behavior, though its source guidance and triggers are broad.
Before installing, understand that this skill is written in Chinese and may answer sensitive historical questions from official-authoritative source framing. It does not appear to access private data or alter your system, but users wanting balanced historiography should ask for multiple perspectives or alternative sources explicitly.
Mandating use of 'official authoritative sources' for sensitive historical topics without offering alternatives or noting potential bias can hard-code a one-sided response policy into the skill. In a historical Q&A context, this can systematically steer outputs toward censored, propagandistic, or incomplete narratives, reducing user autonomy and answer neutrality.
The trigger list is overly broad and includes generic ideological terms plus an open-ended catch-all phrase, which can cause the skill to activate outside its intended scope. This increases the chance of unintended routing, content hijacking, or suppression of more appropriate skills when users mention common political or historical concepts.
The entire skill content is written only in Chinese, and there is no indication that the user can choose another language or that the locale restriction is intentional and documented. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.
The module title and description are written entirely in Chinese and narrowly define the skill as answering questions about Mao Zedong Thought, with no indication that users may choose another language or locale. Under the policy rule, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is explicitly documented and justified.
The function advertised as generating an answer from a user's question ignores the question entirely and returns the full knowledge base. In an agent skill, this can cause unintended disclosure of all embedded content, bypass least-privilege expectations, and enable prompt/context injection if the knowledge file contains adversarial or sensitive text.
No suspicious patterns detected.
