Shobr: Job seach CLI via browser automation, event-sourcing, LLM

Introducing Shobr, a stealthy and UNIX-iest job search automator that utilizes browser automation, event-sourcing, and large language models. This innovative tool is designed to simplify and enhance the job search process in 2026's competitive market. The key requirement for Shobr is the beachpatrol browser, which allows for authentic scraping of job listings while avoiding detection by mimicking regular user behavior.

Shobr streamlines the job search process through five pipeline stages, starting with LinkedIn job search result scraping based on user-defined keywords and locations. It then visits individual job pages to extract full descriptions, salary ranges, and Easy Apply links, ensuring that requests are paced to avoid rate limits. Shobr's scoring system compares enriched jobs against the user's personal Markdown profile and deal-breakers, and it leverages an LLM to rewrite resumes to highlight relevant skills.

The CV toolchain ensures that the rewritten resume fits perfectly on one page, with local Kanban-style tracking for job applications. All user data is stored under $XDG_CONFIG_HOME/shobr/, and the beachpatrol browser profile holds the logged-in LinkedIn session. Job titles are fed to LinkedIn searches as ORed keyword queries, and employment type filters are applied.

The tool also allows users to configure geo targets and reject employment types as needed. Shobr's event-sourcing pattern ensures that every pipeline stage is logged in an append-only events.jsonl file, which is replayed to create a .json projection of the current state.