Security audit
Security checks for vulnerabilities and agentic risk
This is a static knowledge-base skill about agent-first engineering; the noted issues are documentation clarity concerns, not evidence of unsafe behavior.
Before installing, treat this as advisory engineering guidance rather than authoritative policy. If you use the maintenance checklist, only fetch external sources and update repository files when you explicitly intend that work and have reviewed the sources.
The README states the skill 'loads automatically' for broad topic areas, but it does not define precise trigger boundaries, precedence, or opt-in behavior. In an agent toolchain, ambiguous auto-activation can cause the skill to be invoked in unintended contexts, injecting irrelevant or conflicting guidance into prompts and increasing the chance of unsafe or erroneous agent behavior.
该段说明 skill 会在一组较宽泛的话题下“自动加载”，但没有定义更明确的触发边界、优先级或排除条件。结果可能是在仅部分相关的对话中被意外激活，向代理注入不必要的上下文或行为预期，增加误用、越权建议或提示词污染面的风险。
The source URL is fixed to the zh-Hans-CN locale and the main instructional content begins in Chinese, with no indication that users may choose another language or that the skill is intended only for Chinese-speaking contexts. This creates a natural-language locale policy issue because the skill implicitly enforces a specific language presentation without opt-in.
This markdown file presents the skill content entirely in Chinese, starting with a Chinese title and continuing throughout, without offering a language choice or explaining a region-specific requirement. The stated policy flags language or locale constraints when they are imposed without user opt-in or clear justification.
The manifest describes the skill as a knowledge base distilled from OpenAI's Harness Engineering post, which implies a bounded source basis centered on that post. This file explicitly says the skill also integrates broad industry best practices from many external sources, materially expanding the skill beyond the manifest-described basis.
The maintenance workflow explicitly instructs use of web_fetch to retrieve third-party content and then update repository files, which describes an agentic operational loop rather than static reference material. If an agent follows this embedded workflow without strict authorization and source-validation controls, it could ingest untrusted content or make repository modifications beyond the user’s intent.
The title and cited source URL indicate the content is presented specifically in Simplified Chinese (zh-Hans-CN), with no note that other languages are available or that the locale is optional. This is a natural-language policy concern because the file implicitly fixes the skill content to one language without user opt-in or documented regional justification.
The file prominently links to a Chinese documentation version, but there is no accompanying statement about language selection behavior or whether users can choose their preferred language. In multilingual skill ecosystems, language behavior should be explicit to avoid implicit locale assumptions.
该文件在目录说明中将自身标注为“中文版介绍”，整体内容也默认以中文呈现，但未在文内说明这是用户自选语言或提供语言选择机制。根据规则，若技能在自然语言层面强制特定语言而无用户选择或明确合理性说明，可构成语言策略问题。
This markdown file contains guidance presented in Chinese with embedded English slogans, but it does not indicate whether the language choice is intentional, optional, or tied to a specific audience. Under the language/locale policy, forcing a specific language without user opt-in can be a natural-language policy issue.
No suspicious patterns detected.
