Open-source job search tool ai-job-search has reached the top spot on GitHub’s trending list this week. The project lays out a Claude Code-based agent workflow that can screen job openings, rewrite resumes, and prepare users for interviews.
According to BlockTempo, the project’s creator, Mads Lorentzen, was previously a geophysicist. After his position was eliminated at the end of 2025, he used the workflow to send 69 tailored applications, landed 20 first-round interviews, and signed one contract. He started his new role as an AI engineer in June 2026.
The repository has accumulated about 61,600 stars and 19,000 forks on GitHub so far.
The workflow covers several parts of the hiring process
ai-job-search runs through a set of slash commands. The /setup command reads materials such as a resume PDF and LinkedIn export files to build a personal profile. /scrape searches job sites and sorts openings by fit, while /rank scores listings in batches and produces a priority list.
The centerpiece is /apply. It first writes a tailored resume and cover letter, then hands them to a second AI that has not seen the draft and is tasked with finding problems. After revisions, the materials are compiled into PDF format, and the system checks both layout and whether an applicant tracking system, or ATS, can read the files properly.
For interview prep, /interview creates preparation packs for different interview stages and runs mock interviews. The rule set is explicit: if there is a gap in a candidate’s background, the answer should bridge it honestly and should not invent experience.
The /outcome followup function drafts follow-up emails for applications that have gone quiet for a period of time, with a default setting of 10 days. It only prepares drafts, does not send them, and caps follow-ups at two per application.
Creator says he disclosed his use of the tool
Lorentzen said he told every company he interviewed with that he had used the workflow. He said that did not count against him and often turned into a technical discussion during interviews.
Cost and market coverage remain practical constraints
Claude Code does not offer a free plan. Users need a Claude Pro, Max, or Team subscription, or they can switch to usage-based billing through the Anthropic API.
People using other tools such as Codex or Gemini CLI can use the job-search component directly, but the full workflow still depends on community modifications.
Out of the box, the project supports four Danish job sites, LinkedIn public listings, and the tech jobs aggregation site freehire.me. For other markets, users need to generate their own search tools through /add-portal. Sites that require login are rejected.
For job seekers in Taiwan, core functions such as building a personal profile, evaluating fit, and writing resumes and cover letters are not tied to a specific country or language. The job-search portion, though, still needs AI-side optimization. A list compiled in the project’s discussion area includes versions for China, South Korea, Malaysia, India, and the UK, but there is currently no connection to Taiwan job boards.
Privacy is another issue users need to watch
The project also raises privacy concerns. GitHub does not allow public repositories to be forked into private ones, while /setup writes a user’s name, contact details, work history, and salary expectations into files tracked by Git.
Lorentzen’s recommendation is to create a private repository if the tool is only being used for a personal job search, and to fork the project only when planning to contribute code.


