Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Shopee market-analysis skill that uses a disclosed Qinghu API and local exports, with no executable payload or hidden behavior found.
Install this if you intend to use Qinghu’s Shopee data APIs and are comfortable with API calls that may consume Qinghu points. Before using it, confirm whether the agent may use QINGHU_TOKEN or QHKIT_TOKEN from your environment, and expect larger result sets to be exported as local files.
The metadata description says the skill 'must trigger' for broad Shopee-related topics, which can cause over-selection of this skill even when the user did not ask to use external paid tooling or export data. In an agent environment, overly broad mandatory triggers increase the chance of unintended API calls, unnecessary credential prompts, and exposure of commerce data handling paths.
The skill allows obtaining the API key from environment variables without an explicit user-facing warning or permission check. In agent systems, silent access to ambient credentials is sensitive because it can cause the model to use secrets the user did not intend for this task, and may normalize hidden credential consumption for paid third-party APIs.
L047 将工具字段说明写为“title（中文名）”，整体技能说明也默认以中文交互，没有看到允许按用户偏好切换语言或明确的中文-only 合理性说明。根据规则，若技能强制特定语言而无用户选择，可能构成自然语言策略问题。
The skill instructs the agent to automatically export large result sets to a local file and reuse cached file paths, without requiring explicit user consent for data persistence. This creates a data-handling risk because marketplace records, user-specified queries, or derived commercial analysis may be written to local storage unexpectedly and retained beyond the immediate response.
The '何时触发' examples only show positive matches and omit nearby cases that should stay in normal chat flow. That ambiguity can cause accidental activation for loosely related requests, leading to unnecessary external calls or prompting for secrets when simpler, non-tool responses would suffice.
