kunchen1110/interviewradar

InterviewRadar is a Python-based workflow that generates a personalized Chinese interview preparation package from your resume and a rough description of the job type you are targeting. It is designed to be run through Claude Code or OpenAI Codex as a skill, rather than as a standalone application. The tool scrapes real interview posts from platforms such as NowCoder (牛客), GitHub, and public tech blogs, then applies a 730-day recency filter to keep only recent questions. It weights questions by how often they appear and how recent they are. After collecting and ranking questions, it maps each high-frequency question to the specific projects listed on your resume to produce a chain of follow-up questions, rather than a generic list. The output is a Markdown document written in Chinese. A key design goal is traceability: every question in the output includes a link to the original post where it appeared in a real interview. This is different from static question banks, which become outdated, and from AI-generated questions, which may not reflect what interviewers actually ask. The tool supports any job category because it generates its search terms from your resume and job description rather than relying on a preset topic list. An optional Xiaohongshu (Little Red Book) data source can be added by setting up a separate tool called MediaCrawler. That platform requires cookie-based authentication and cannot be scraped directly. When configured, the agent calls MediaCrawler automatically for that source and uses OCR to extract question content from images in posts, since many interview posts on that platform put key content in pictures rather than text. The tool is licensed under MIT.