How an Open‑Source Multi‑Agent AI System Transforms Job Hunting

Problem Overview

Job seekers face five major pain points: low‑efficiency mass applications, ATS filtering failures, chaotic workflow management, lack of data‑driven decision making, and costly manual resume customization. These issues waste valuable time, especially for technical professionals.

Project Overview

Career‑Ops (v1.2.0, released 2026‑04‑08) is an open‑source, multi‑agent job‑search automation system built on Claude Code. It converts an AI‑coded CLI into a full‑process job‑search command center, replacing manual spreadsheet management with a precise, engineered workflow.

Technical Architecture

Core Agent Stack

The system centers on a Claude Code agent augmented with 14 custom skill modes that parse job descriptions, evaluate resume fit, generate interview material, and draft negotiation scripts. Parallel sub‑agents handle batch job processing, keeping the workflow efficient.

Job Evaluation Logic

Job‑type identification (LLMOps, Agentic, PM, SA, FDE, Transformation, etc.)

Quantitative scoring using an A‑F grade combined with a 5‑point scale across ten weighted dimensions (match quality, salary, career growth, workload, etc.)

Deep matching analysis that highlights skill gaps and rank suitability, recommending a minimum score of 4.0/5 before applying.

Resume Generation

HTML templates rendered with Playwright produce ATS‑compatible PDFs, automatically injecting job‑specific keywords while preserving visual design.

Data Management & Visualization

Data is stored locally in Markdown, YAML, and TSV files (git‑ignored for privacy). A TUI dashboard built with Go, Bubble Tea, and Lipgloss provides six filter tags, four sorting modes, and both grouped and flat views of the application pipeline.

Technology Stack

Agents: Claude Code (custom skills & modes)

Automation: Playwright (web scraping, PDF generation)

Backend Scheduler: Node.js

Dashboard: Go + Bubble Tea + Lipgloss (Catppuccin Mocha theme)

Configuration & Storage: YAML, Markdown, TSV

Usage Workflow (Simplified)

Paste a job URL or JD description.

System auto‑detects the job prototype.

Six‑module deep evaluation + quantitative scoring.

Generate evaluation report, ATS‑ready resume, and progress record.

Track status in the terminal dashboard.

Human reviewer decides whether to submit, completing an AI‑assist + human‑in‑the‑loop cycle.

Quick‑Start Guide

Prerequisites: Node.js, Go (for the dashboard), and a functional Git/terminal environment.

# 1. Clone the repository and install dependencies
git clone https://github.com/santifer/career-ops.git
cd career-ops && npm install

# 2. Install Chromium (required for PDF generation)
npx playwright install chromium

# 3. Verify environment
npm run doctor

# 4. Configure personal profile and target companies
cp config/profile.example.yml config/profile.yml   # edit with personal info
cp templates/portals.example.yml portals.yml      # customize target firms

# 5. Prepare your resume (cv.md in Markdown)
# 6. Launch the Claude Code agent
claude

# 7. Issue custom commands, e.g.:
# "Change job prototype to backend engineer"
# "Set salary floor to 20K"
# "Add 5 target companies to portals.yml"

# 8. Start using the system:
#   • Paste a job URL for full automation
#   • Or run specific sub‑commands

Core Commands

/career-ops               → List all available commands
/career-ops {JD/URL}      → Full automation (evaluation + PDF resume + tracking)
/career-ops scan          → Auto‑scan target companies for new jobs
/career-ops pdf           → Generate ATS‑optimized resume only
/career-ops batch         → Parallel evaluation of 10+ jobs
/career-ops tracker       → View all application progress in the terminal
/career-ops deep          → Deep research on target companies
/career-ops apply         → AI‑assisted form filling

Dashboard Launch

cd dashboard
go build -o career-dashboard .
./career-dashboard --path ..   # start visual tracker

Best Practices & Tips

Initially enrich your cv.md with detailed achievements and clear job preferences to improve evaluation accuracy.

Always keep a human in the loop; AI suggestions must be manually verified before submission.

Install the Chromium core and run npm run doctor to avoid runtime errors.

Use the standard GitHub URL when cloning to prevent “link dead” errors.

Leverage Claude commands to retarget the system for different roles (e.g., switch from backend to AI‑research).

Regularly pull updates from the repository to benefit from new features and community contributions.

Long‑Term Value

Beyond job hunting, the evaluation framework, story‑bank, and pipeline management can be repurposed for career planning, skill‑development tracking, or assessing the value of training programs and personal projects.

https://github.com/santifer/career-ops