Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese-language prompt toolkit for analysis and writing, with some usability and trigger-scope caveats but no evidence of hidden execution, data access, persistence, or exfiltration.
Install this if you want Chinese-language prompt templates for structured thinking and official-style writing. Review or narrow the keyword triggers if accidental activation would be disruptive, and be aware that the public-document prompt includes vague political/sensitive-word constraints that may shape output style.
The skill is written entirely around Chinese trigger phrases and includes a tool explicitly named 汉语新解, but nowhere states that Chinese-only behavior is optional or limited to a justified region-specific use case. This can amount to a language/locale policy issue because users are not offered an opt-in or alternative language behavior.
The top-level trigger list includes broad everyday phrases such as '抽象', '多角度', '定义', and '打分'-like concepts that can easily appear in normal user requests. This creates ambiguous activation conditions, increasing the chance the skill is invoked unintentionally and alters assistant behavior when the user did not explicitly request this skill.
Per-tool trigger phrases like '追问', '本质', '定义', '探索', and similar everyday expressions are vague and unconstrained, so regular conversation may match them accidentally. That can cause tool misfires, unexpected prompt steering, and unreliable behavior across unrelated tasks.
The trigger list contains short, generic keywords such as "定义", "矩阵", "抽象", and "公文" that are common in ordinary Chinese user requests. With keyword activation mode, these broad matches can cause the skill to activate unintentionally, leading to prompt injection of this skill's instructions into unrelated conversations and reducing user control over which tool is invoked.
This markdown skill defines its role, goals, scoring dimensions, and output format entirely in Chinese, which effectively forces a specific language/locale for use. The policy allows fixed locale behavior only when the skill offers opt-in or clearly documents a justified regional constraint, neither of which appears here.
The skill explicitly constrains output to be 'politically correct' and to avoid 'sensitive words' without any user opt-in, policy citation, or documented justification. This creates a covert behavioral restriction that can bias, suppress, or reshape user-requested content in ways that are not transparent, which is especially risky in a system prompt because it silently governs all downstream outputs.
The skill is written entirely in Chinese and prescribes a Chinese-only response format without any user opt-in or fallback. This can override user language preferences, reduce usability, and create safety/comprehension risks for users who cannot read the output, though it is not inherently malicious.
The prompt is entirely written in Chinese and implicitly expects Chinese-language interaction, but there is no documented locale constraint or user opt-in. This can cause the agent to ignore user language preferences, reduce accessibility, and produce unusable responses for non-Chinese-speaking users, which is a genuine prompt-safety and usability issue even though it is not overtly malicious.
The entire prompt is written as a mandatory Chinese-language system prompt and provides no indication that the user may choose another language or that the language restriction is optional. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.
The constraint at L30 says '只输出数据库已有信息' ('only output information already in the database'), which implies no new reasoning or derived content should be produced. However, the prompt explicitly instructs the model to extract implicit premises, diagnose logical flaws, and perform logical deduction, all of which require generating analytical conclusions not literally present in the input/database.
The prompt is written entirely in Chinese and includes rigid output constraints that implicitly require responses in Chinese, without offering the user a language choice or documenting a justified locale restriction. This can violate language/locale policy when users have not opted into Chinese output.
The entire prompt is written as an instruction set in Chinese and specifies the output format in Chinese, with no indication that users may choose another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is clearly justified.
Natural-language fields including the description and trigger phrases are Chinese-only, which can imply a fixed language experience without documenting opt-in or a region-specific constraint. Because no language choice or justification is provided in the manifest, this may conflict with language/locale policy expectations.
The entire prompt is written as prescriptive Chinese-language instructions and the required output template is also fixed in Chinese. This can violate a language/locale policy when the skill does not give users any option to choose another language or explicitly opt into Chinese output.
No suspicious patterns detected.
