Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed workflow helper for clarifying routine code-change requests and deciding when to ask for confirmation, with no bundled executable code or hidden persistence.
Install this if you want a Chinese-language workflow that asks clarifying questions before many code changes. Review the auto-execution rules first, especially single-file edits, text-only edits, renames, and 'directly modify' requests, and tighten them if you prefer confirmation before every change.
The manifest description is entirely in Chinese and presents the skill behavior as Chinese-localized by default, with no indication that the user can opt into another language. Under the stated policy, locale-specific behavior should either be optional or explicitly justified as region-specific.
The activation examples include generic phrases like "优化一下", "重构", "改进", and "帮我改改这个文件", plus the catch-all "任何日常代码任务". These descriptions do not clearly bound when the skill should activate versus when another skill or normal handling should apply, increasing the risk of unintended invocation.
The skill explicitly supports autonomous execution when certain heuristics are met, including text-only and single-file changes. In isolation that may seem convenient, but for a code-modifying agent it creates a real risk of the system making consequential changes without adequate user confirmation, especially because the safety classification depends on shallow rule matching rather than robust semantic validation.
- "改进一下"
# 自动执行规则（满足任一即自动执行）
auto_execute:
text_only: true # 纯文本修改
single_file: true # 单文件小改动
safe_operations: # 安全操作关键词
The auto-execution keyword list treats broad, common requests such as '注释', '格式化', '重命名', and especially an explicit '直接改' signal as sufficient to proceed automatically. In a requirement-clarification skill, this can cause the agent to skip meaningful scope validation and act on ambiguous user intent, increasing the chance of unauthorized or unsafe modifications.
Keywords like "替换" and "改掉" are broad action words that can describe many routine edits, but the file does not specify what kinds of replacements or changes require confirmation versus those that do not. This ambiguity can lead to inconsistent or unintended confirmation behavior.
该 markdown 文件适用 SQP-1。L021-L022 列出的默认触发词如“感觉”“可能”“差不多”“改进”等属于常见日常表达，缺少上下文约束或排除条件，容易与普通交流重叠并触发追问规则。
The document defines an auto_execute mode where satisfying any one condition causes execution without confirmation. In a code-modification agent, this increases the chance of unintended or insufficiently reviewed changes, especially when the criteria include subjective notions like 'safe operations' or user phrasing such as 'directly modify'.
默认触发关键词：
`大概`、`优化一下`、`类似`、`感觉`、`可能`、`差不多`、`改进`
### auto_execute（自动执行规则）
满足任一条件即自动执行，无需确认。
The recommendation to 'reduce confirmations' by expanding auto_execute.safe_operations encourages broader unaudited execution. In this skill context, even seemingly simple actions like renaming, reorganizing code, or adding logging can alter behavior, leak information, or degrade stability if performed automatically.
如果希望更少确认：
auto_execute:
single_file: true # 保持
text_only: true # 保持
# 添加更多安全操作
The skill content is entirely written to drive Chinese-language interaction and examples without any indication that the user requested or consented to that language. In an agent skill, forcing a language can degrade user comprehension, cause incorrect confirmations or approvals, and create a misleading user experience, especially in requirement-clarification workflows where precise understanding matters.
SQP-3 适用于所有文件。该文档从标题到内容均以中文呈现，未说明是否支持其他语言，也未提供用户可选择语言/locale 的选项，可能构成未获用户选择的语言策略限制。
No suspicious patterns detected.