Security audit
Security checks for vulnerabilities and agentic risk
This skill is a Chinese-language case-matching assistant that reads its bundled case database and can save a Markdown report, with no evidence of hidden execution, exfiltration, or system persistence.
Installers should expect a Chinese-language workflow that reads the included case database, asks about their business situation, and may save a local Markdown report. Consider asking the agent to show the report inline or confirm the filename before writing if you do not want files created automatically.
The trigger phrases are broad enough to match ordinary conversation, which can cause the skill to activate unexpectedly and access local case data or start a structured workflow without clear user intent. In agent environments, unintended invocation increases the chance of surprising file reads and subsequent actions, especially when combined with default file writing behavior.
The skill instructs the agent to write a Markdown file by default without obtaining user confirmation, creating an unannounced side effect on the filesystem. Even though the content is not inherently sensitive, automatic writes can surprise users, overwrite expectations, leak private prompts into saved artifacts, or be chained with broad triggering to create unwanted files.
The default prompt contains a broad skill invocation token ('$case-explorer') without any activation constraints, scope checks, or user-confirmation language. This can cause the skill to be invoked in contexts the user did not clearly intend, increasing the risk of accidental routing, prompt confusion, or misuse by downstream agents that treat the trigger text as authoritative.
This JSON file stores all human-readable case metadata exclusively in Chinese, including titles, chapters, scenarios, and takeaways. If this file is used as a skill reference corpus, it effectively constrains outputs or interactions to a specific language without any documented user choice or opt-in.
All user-facing instructions, triggers, workflow steps, and output templates are presented only in Chinese, effectively constraining interaction to a specific language. The file does not indicate that users may choose another language or that the Chinese-only behavior is a documented, justified locale limitation.
The manifest description frames the skill as helping readers match relevant book cases and analyze success factors, which is primarily an analysis/advisory function. The SKILL instructions additionally direct the agent to persist results to a Markdown file by default, which is a write side effect not stated in the manifest-level description.
The file presents the skill name and description in Chinese while the default prompt is fixed in English, which can impose a language choice without explicit user opt-in. This may conflict with language or locale policy expectations when the skill does not state that users can choose their preferred language.
