Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese Gaokao essay-coaching skill with local reference files and simple local scripts; its broad trigger wording is worth noting, but the inspected artifacts do not show harmful behavior.
Install this only if you want Chinese Gaokao essay coaching. Expect Chinese output, use of bundled scoring and writing references, and optional reading of essay text or an essay file you provide; avoid pointing the --file option at sensitive non-essay files. Be aware that broad trigger terms may make the skill activate in some general Chinese writing conversations.
描述强调的是覆盖写作全过程的作文指导能力，包括写前的审题立意、结构搭建、素材推荐，以及写后的润色、评分与诊断。但代码只实现了对已提供作文成品的分析与评分预测：读取文本/文件后，检查标题、字数、段落、应用文格式、语言表达，并给出预估分数和改进建议。它没有任何针对题目材料的审题立意分析、提纲生成、素材库推荐、范文参考输出或文本润色改写功能。因此，实际能力只是声明中的“评分预测与诊断改进”子集，无法支撑“全流程指导”这一核心定位，属于实质性描述不符。
该描述将技能定义为覆盖高考作文写作全流程的综合指导工具，但代码只实现了其中很窄的一部分：按预设模板生成作文结构框架，并附带少量固定写作提示与应用文格式。它不分析题意、不推荐素材、不进行语言修改、不打分、不诊断问题，也不提供范文。因此声明的能力范围显著大于实际行为，属于描述与实现不一致。虽然“结构搭建”这一项与代码相符，但不足以支撑“全流程指导”的整体声明。
The skill explicitly instructs the agent to read local reference files before responding, but it does not declare any tool scope or allowed-tools boundary. That mismatch creates an undeclared file-read capability, which weakens least-privilege controls and can lead to unauthorized or surprising access to workspace files if the skill is invoked in broader contexts.
The trigger phrases are broad enough to match ordinary writing-related discussion, causing the skill to activate when the user may not want or expect it. Unintended invocation is a security-relevant control issue because it can expand access to file-reading behavior and alter agent behavior outside the intended task boundary.
Using an open-ended trigger list with “等” leaves the activation boundary undefined, making it difficult to predict when the skill should run. Ambiguous routing increases the chance of accidental activation and unintended use of the skill’s file-reading workflow in unrelated conversations.
The text states that language must follow modern Chinese norms and bans internet slang or coined words, which imposes a specific language/locale style requirement. Because this file does not present that constraint as an optional, user-selected setting or clearly justify it as a region-specific limitation, it creates a natural-language policy concern.
This markdown file contains natural-language instructions that require outputs to use specific Chinese terminology and later prescribes conversion of expressions into formal written Chinese. Because the policy category applies to all file types and forbids forcing a specific language or locale without user opt-in, these mandatory language constraints are a policy concern.
These lines instruct that colloquial wording 'must' be identified and changed into written Chinese, and provide rigid replacement rules. That is a natural-language policy issue because it imposes a specific language/register choice rather than allowing the user to select their preferred style.
The file presents itself entirely in Chinese and is specifically framed as a 高考作文 tool, while the CLI descriptions and all user-facing output are also fixed to Chinese. This enforces a single language/locale experience without any user opt-in or alternative, which matches the policy category for language or locale constraints.
The module title, function description, and CLI usage are all presented exclusively in Chinese, and the supported argument values are also fixed to Chinese labels. This imposes a language/locale constraint on users without any opt-in, fallback, or explanation that the skill is intentionally region-specific.
This reference file presents all descriptions and instructional content only in Chinese, and does not mention that the skill is Chinese-language only or offer an alternative language choice. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.
