Security audit
Security checks for vulnerabilities and agentic risk
The skill is not malicious, but it gives broad, under-scoped instructions to monitor external sources and create or update skills without clear user control.
Review before installing. The package does not show malware-like behavior, but its instructions are broad: it may lead an agent to track external sources and create or update skills unless you keep those actions user-approved and scoped. Treat it as a Chinese-language research workflow, not a complete automated Karpathy research implementation.
声明描述的是一个面向特定研究对象的持续研究技能，重点在于系统性分析 Karpathy 的公开作品并提出三次拷问。实际代码只是在内存中记录概念解释、生成三个固定问题，并用随机数给出“理解质量”判断；示例中甚至只分析了“神经网络”这个硬编码概念。代码没有网络请求、文件扫描、项目解析、YouTube/课程内容处理、定时运行或成果转化逻辑。额外的 AdvancedTestSuite 只是压力/边界/性能/集成测试占位实现，与声明目标关系不大。因此描述与实际行为存在明显不匹配。
The manifest description and headline content are written entirely in Chinese and present the skill's behavior as fixed, with no indication that users can opt into another language or locale. Under the stated policy, a skill that effectively enforces a specific language without user choice is a natural-language policy violation unless the locale constraint is clearly justified.
The module title and descriptive text are written exclusively in Chinese, and the method docstrings and generated prompts also assume Chinese-language operation. This indicates a language-specific behavior without any visible opt-in or user choice, which matches the policy concern for forced language or locale constraints.
The manifest describes a skill that studies Karpathy's projects and educational content daily, including specific targets like llm.c, nanoGPT, micrograd, CS231n, and YouTube material. In contrast, the implementation only stores a supplied concept/explanation, generates three generic questions, and returns a random quality score; it never fetches, reads, or analyzes any Karpathy work.
The manifest explicitly claims the skill converts research outcomes into another derived capability, but no code performs transformation, export, synthesis, or generation of any downstream 'Musk skill' artifact. The implementation remains a simple in-memory analyzer/test harness with print statements.
No suspicious patterns detected.
