Security audit
Security checks for vulnerabilities and agentic risk
This JSON formatting skill is mostly purpose-aligned, but it asks for read/write/exec authority while describing overly broad activation and inconsistent data-flow boundaries.
Review this skill carefully before installing. Use it only for explicit JSON formatting, minifying, validation, or path extraction tasks, and avoid letting it run broadly on unrelated code or deployment work. Check any file writes before replacing originals, especially when key sorting or Unicode escaping options are enabled.
The activation guidance is overly broad, covering generic coding, debugging, and deployment situations rather than narrowly scoped JSON-formatting tasks. In agent environments, this can cause the skill to trigger on unrelated developer requests and gain access to read/write/exec capabilities in contexts where it is unnecessary, increasing the chance of unintended command execution or data exposure.
The capability coverage section includes an excessively broad keyword list, including generic development terms, which can make the skill match many unrelated prompts. Because the skill advertises write and exec tools, over-triggering expands the operational surface and can lead to inappropriate invocation in sensitive workflows.
L255 将 LLM API 标为必需外部依赖，而 L259-L260 明确声称无需额外 API Key 且所有处理在本地完成、不依赖外部服务。这不是简单遗漏，而是对技能是否依赖外部服务的直接冲突，会误导用户对执行边界和数据流向的判断。
The skill name, summary, description, and usage content are presented in Chinese throughout, with no indication that users may choose another language or that the skill is intentionally restricted to a Chinese-language audience. This creates a locale/language constraint that is not opt-in or justified in the file.
The skill description explains transformation features but does not clearly warn that outputs may change whitespace, escaping style, or key order, which can affect downstream expectations or diffs. Although the document later mentions sorting and Unicode handling, there is no upfront caution about these potentially impactful changes to user data representation.
L060 明确说免费版验证能力“不修复错误，仅报告”，这是该版本的重要边界；但整份文档多处将“处理输入参数,完成核心逻辑,返回结构化响应”泛化描述为可执行完整处理，且免费版/专业版能力边界直到后文才澄清。虽然不是代码级行为证据，但文档在验证能力的预期上存在混淆，容易让用户误判是否包含修复。
No suspicious patterns detected.
