Security audit
Security checks for vulnerabilities and agentic risk
This is a small Baidu web-search CLI skill that sends an explicit search JSON request to Baidu using a user-provided API key, with no hidden install scripts or persistence.
Install only if you intend to use Baidu's AI search API and are comfortable sending your search terms, filters, and API authorization to Baidu. Edit search_request.json for your own query and avoid including confidential, regulated, or internal-only information in search requests.
The skill explicitly instructs sending user-supplied search content and related parameters to Baidu's remote API using curl, but it does not warn users that their queries leave the local system or may be processed by a third party. This creates a real privacy and data-handling risk, especially if users search sensitive internal, personal, or regulated information.
The curl command performs direct external transmission of request data to a third-party Baidu endpoint, including user query content and authentication context. In the skill context, this is expected functionality for web search, but it is still security-relevant because sensitive data could be exfiltrated off-system if users or upstream agents place confidential content into search_request.json.
4.如果环境变量BAIDU_API_KEY存在但curl返回身份验证错误,返回信息:APIKEY验证失败，请检查你的环境APIKEY是否设置正确
curl -X POST -H "Authorization: Bearer apikey" -H "X-Appbuilder-From: openclaw" -H "Content-Type: application/json" -d "@search_request.json" "https://qianfan.baidubce.com/v2/ai_search/web_search"
The JSON payload fixes the user message content to "中国新闻", which imposes a specific language/locale in a natural-language field. Per the policy, language constraints should not be forced unless the skill offers user choice or clearly documents a justified locale-specific scope.
The instructions and prescribed returned messages are all in Chinese, which can impose a specific language on users without opt-in. Under the natural-language policy rule, forcing a language or locale without user choice is a policy concern unless the skill clearly documents a justified region-specific limitation.
No suspicious patterns detected.