Overview

This is a job-application framework that runs locally on top of Claude Code. After you fill in a profile, Claude can evaluate job postings against your background, rank scraped positions, draft and review a tailored CV and cover letter in LaTeX, compile the PDFs, and build interview preparation material. It targets people running a personal job search who are comfortable with a command-line toolchain and version-controlled repository.

The workflow is organized around /setup, /scrape, /apply, plus additional commands for ranking, interviews, outcome tracking, Gmail/Notion sync, HTML reports, template registration, and new job portals. The core profile, fit evaluation, and drafter-reviewer pipeline are described as language- and country-agnostic; the shipped portal search skills focus on Danish job boards, with linkedin-search and freehire-search as broader starting points. See the command reference for the full list.

Quick Start

Prerequisites are Claude Code with a paid Claude subscription or API credits, Python 3.10+, Bun, and a LaTeX distribution providing lualatex and xelatex. pypdf is optional for the ATS parseability check.

1. Fork and clone the repository, or create a private repository with the original as upstream if the copy will contain your personal data. The README notes that a fork of a public repository is always public, and /setup writes personal details into tracked files.

2. Install the portal CLI tools:

   python3 tools/install_portal_clis.py
   

   On Windows, run python tools\install_portal_clis.py. The installer uses bun install --ignore-scripts and skips packages with lifecycle scripts or related configuration files.

3. Start Claude Code with claude, then run /setup. It can read a populated documents/ folder, import a single CV pasted into chat, or conduct an interview.

4. Run /scrape to search the configured portals for matching postings, or /rank to batch-score a larger scrape result.

5. Apply with a posting URL or pasted job description, for example:

   /apply https://jobindex.dk/job/1234567
   

   The workflow evaluates fit, drafts the CV and cover letter, spawns a reviewer agent, revises, compiles and inspects the PDFs, and checks the CV's extracted text layer for ATS compatibility. The result is presented with a verification checklist.

The setup guide and the project's security notes cover the detailed update and untrusted-posting handling workflows.

Constraints

• Claude Code requires a Claude Pro/Max/Team subscription or Anthropic API credits; there is no free tier.
• CV generation uses lualatex, while the cover letter uses xelatex. A minimal TeX installation may need extra packages.
• The shipped Danish portal skills are market-specific. Other markets are expected to use linkedin-search, freehire-search, or a portal skill generated with /add-portal. LinkedIn access is intended for personal use only and requires keeping request volume low.
• /setup writes personal data into repository files, so the project recommends a private repository rather than a public fork for non-contribution use.
• Job postings are treated as untrusted input, but the README states that the agentic defenses are instruction-level rather than a sandbox.
• The repository is MIT-licensed.

HotRepo's take

No HotRepo daily ranking, snapshot trend, or Hacker News discussion record is available for this repository, so the site's data does not support an observation about its popularity trajectory. The repository metadata shows a very large star and fork count relative to its creation in March 2026, but without a matching trend record those numbers cannot be tied to a specific ranking or growth window.

What the material does show is a fairly complete, inspectable workflow: six core job-portal CLI skills, a drafter-reviewer /apply pipeline with PDF layout and text-layer verification, template and portal extension points, and explicit handling of personal data and untrusted postings. The README also includes a personal hiring anecdote as evidence that the workflow was used; that is the author's account, not an independently measured outcome.