What is job-searcher?

job-searcher is a Claude Code skill pack that collects public job postings from 22 Korean and international job platforms and normalizes them into a single JSON format. The idea is to let an AI agent handle the repetitive task of checking multiple job boards like JobKorea, Saramin, Wanted, and Jumpit, rather than a person visiting each site manually. It also normalizes each posting's deadline into a status of open, closing soon, closed, or rolling, so a user can quickly filter to postings that are still accepting applications. The tool is explicitly read only. It does not log in anywhere, does not save data on its own, and does not submit job applications. It only fetches public search, listing, or RSS pages using GET requests. The recommended way to use it is as a Claude Code plugin, after which a user can ask in plain language for things like postings for a specific role from specific sites, and the agent collects them. It can also be run directly as Python scripts per platform, each of which is self contained with its own fetching, parsing, and output logic. Of the 22 supported platforms, 21 were verified live and working as of the README's writing, each returning real postings through public pages or APIs. One platform, RocketPunch, uses a JavaScript-only page with anti-bot protection, so its collector honestly reports that a browser is needed rather than fabricating data. The project states a design principle of honesty: any field a source does not expose is left null rather than guessed, and a collected posting is not guaranteed to still be open. Installing requires the Python packages curl_cffi, beautifulsoup4, and lxml, which the agent installs automatically when used as a plugin. The project is MIT licensed.