AI Job Search is a structured workflow that transforms Claude Code into a personal career assistant. Instead of writing generic cover letters and hoping for the best, you fork the repo, runDocumentation Index
Fetch the complete documentation index at: https://mintlify.com/SudharakaA/ai-job-search/llms.txt
Use this file to discover all available pages before exploring further.
/setup to build a detailed professional profile, and then let Claude evaluate job postings, tailor your CV in LaTeX, write targeted cover letters, and even prepare you for interviews. The whole pipeline runs inside your local Claude Code session, keeping your career data in your own repository.
How the framework works
The workflow has three sequential phases that take you from a blank fork to a submitted application.Profile setup
Run
/setup inside Claude Code to populate your candidate profile. You can point Claude at a documents/ folder of CVs, LinkedIn exports, diplomas, and reference letters; import a single CV by pasting it; or answer structured interview questions. All three paths produce the same output: a set of profile skill files that the rest of the workflow depends on.Job search
Run
/scrape to search multiple job portals in parallel. Claude deduplicates results and ranks them by fit against your profile, then presents the shortlist so you can pick a posting to act on.The drafter-reviewer pipeline
The/apply command deliberately separates writing from critique. A drafter agent tailors your CV and cover letter for the specific posting. A reviewer agent, spawned with a fresh context, researches the company and critiques both drafts for missed keywords, weak framing, and generic language. The drafter then revises based on that feedback.
After revision, Claude compiles each LaTeX document to PDF and visually inspects the rendered pages. The CV must land at exactly two pages with no orphaned entry titles; the cover letter must fit on exactly one page with the signature visible. If layout issues are found, Claude applies targeted fixes (\needspace, \enlargethispage, font-matching wrappers) and recompiles until both documents pass. An optional ATS check then extracts the CV’s text layer with pdftotext and verifies that contact details, reading order, and keyword coverage all look correct to an applicant-tracking system parser.
All claims in the CV and cover letter are verified against your actual profile. The framework will never fabricate skills or experience, and genuine skill gaps are left visible rather than stuffed with keywords.
Language and country support
The core workflow (profile setup, fit evaluation, and the drafter-reviewer application pipeline) is fully language- and country-agnostic. You can target roles in any language and any market. The bundled job search tools are built for the Danish market: Jobindex, Jobnet, Akademikernes Jobbank, and Jobdanmark. If you are in a different country, run/add-portal to generate an equivalent search skill for your local job board. The linkedin-search tool is included as a country-agnostic starting point and works with any location flag (for example, -l "Berlin, Germany" or -l "Remote").
Prerequisites overview
You need four things installed before starting:- Claude Code: Anthropic’s CLI for running Claude in your terminal
- Python 3.10+: required for the salary benchmarking tool
- Bun: the TypeScript runtime for the job search CLI tools
- A LaTeX distribution: TeX Live, MacTeX, or MiKTeX with
lualatexandxelatex
pdftotext from the poppler library for the ATS parseability check. Without it, /apply falls back to a visual keyword review and everything else works normally.
See the Setup guide for platform-specific installation commands.
Key features
Drafter-reviewer pipeline
A second Claude agent critiques every draft before you see it, catching missed keywords and generic language that a single pass leaves in.
PDF verification loop
Every application is compiled to PDF and visually inspected. Page breaks, orphaned titles, and font issues are fixed automatically before output.
ATS parseability check
The CV’s text layer is extracted with
pdftotext and verified against what an applicant-tracking system actually reads, not what the rendered page looks like./upskill command
Analyzes the gap between your profile and your tracked job postings, then produces a prioritized heatmap of skill gaps with web-searched study resources and time estimates.
/expand command
Enriches your profile by scanning public sources you have linked (GitHub, portfolio sites, Kaggle, Google Scholar) and adding discovered competencies with a source tag.
/add-portal command
Generates a job-portal search skill for any public job board in your market, scaffolded from the same structure as the bundled Danish tools.
Get started
Quickstart
Fork the repo, install dependencies, and run your first application in minutes.
Setup guide
Configure your profile in detail across all three
/setup paths.Core workflow
Deep dive into
/scrape, /apply, and the full drafter-reviewer pipeline.