Codex support

[newway-anshul]

How can we use it with codex

[Mechagodz-glitch]

Overview of the career‑ops repository

The open‑source career‑ops repository by santifer provides an AI‑assisted job‑search and application pipeline. It is built around an AI agent (originally Claude) and uses Playwright to automate web interactions. The system focuses on evaluating job postings, tailoring résumés and cover letters to the job description and tracking application progress rather than mass‑applying. Key components include:

CV and profile configuration – Users supply a markdown CV (cv.md) and a YAML profile (config/profile.yml). These files hold personal data such as your summary, experience, targeted roles and salary expectations. The system distinguishes between a user layer (personal files like CV and profile) and a system layer (shared mode files and scripts). Personalisation always goes into the user layer so system updates don’t overwrite your data.
Evaluation and scoring – A job description or posting is evaluated with a six‑block (A–F) scoring rubric. Blocks score the role’s alignment with your archetypes, skills/experience match, compensation, work‑life balance, growth opportunities and company reputation. The pipeline outputs a report in reports/ summarising the score and recommending whether to apply.
ATS‑optimised CV and PDF generation – Scripts generate tailored résumés. Given your CV markdown and the job description, the system uses the AI agent to rewrite bullet points and highlight relevant achievements. It produces either an HTML‑to‑PDF or LaTeX PDF using provided templates.
Portal scanning and batch processing – The scan.mjs script queries ATS portals (Greenhouse, Ashby, Lever, etc.) using zero LLM tokens to fetch new job postings that match your keywords. The pipeline can process multiple postings in parallel and supports batch operations such as scanning, evaluating, PDF generation and updating the application tracker.
Tracking and follow‑up – The system maintains data/applications.md to track applications and their status. Other scripts analyse rejection patterns and help plan follow‑ups.
Human‑in‑the‑loop design – By design, the tool does not auto‑submit applications. It assists with evaluation, tailoring and tracking but relies on you to confirm decisions and send final applications.
Installing and using career‑ops locally
Clone and install dependencies. Clone the repository and run npm install. Install Playwright
for web automation with npx playwright install.
Prepare your data. Create a cv.md containing your résumé in clean markdown (sections such as Summary, Experience, Education and Skills). Copy config/profile.example.yml to config/profile.yml and fill in your name, email, location, targeted roles and narrative. Copy templates/portals.example.yml to portals.yml and specify the companies or keywords you want to scan.
Run the AI agent. Historically, career‑ops was designed for Claude Code. You open a session in your terminal, point the agent at the repository folder and run commands like /career‑ops, /career‑ops‑evaluate or /career‑ops‑scan. The agent reads the mode files and decides how to process the job description. All customisations live in the user layer (CV, profile and modes/_profile.md). The system layer scripts should not be edited directly.
Using career‑ops with a ChatGPT subscription

The repository doesn’t currently include a native ChatGPT workflow. Instead, you can use ChatGPT through OpenCode, an open‑source AI coding agent that supports multiple LLM providers.

Why OpenCode?

The OpenCode project allows you to connect any model from any provider – Claude, GPT, Gemini and more. Its documentation notes that you can log in with your OpenAI account to use your ChatGPT Plus or Pro subscription inside the agent. The career‑ops repository provides a set of slash commands tailored for OpenCode; these commands map to the same tasks defined for Claude, so you can run the pipeline using ChatGPT as the underlying model.

Steps to integrate career‑ops with ChatGPT via OpenCode
Install OpenCode. Follow the installation instructions at opencode.ai
(e.g., curl -fsSL https://opencode.ai/install | bash). You can use the terminal interface or desktop app.
Connect ChatGPT. In OpenCode, run the /connect command and select the OpenAI provider. Log in with your ChatGPT Plus or Pro credentials or provide your OpenAI API key. This step links your ChatGPT subscription to the OpenCode agent.
Open the career‑ops project. In a terminal, cd into the cloned career‑ops directory and start OpenCode there. OpenCode automatically loads context files like CLAUDE.md and modes/*, which describe the job‑search workflow.
Run career‑ops commands. OpenCode exposes a series of slash commands for the job‑search pipeline. According to the repository’s CLAUDE.md, commands such as /career‑ops, /career‑ops‑evaluate, /career‑ops‑scan, /career‑ops‑pdf and others are defined in .opencode/commands/. These commands map to the same tasks as the Claude implementation:
/career‑ops – Show the main menu or evaluate a job description.
/career‑ops‑pipeline – Process pending job URLs in your inbox.
/career‑ops‑evaluate – Evaluate a job offer using the A–F scoring system.
/career‑ops‑compare – Compare and rank multiple offers.
/career‑ops‑pdf – Generate an ATS‑optimised PDF résumé.
/career‑ops‑scan – Scan job portals for new postings.
/career‑ops‑batch – Batch process job offers in parallel.
/career‑ops‑apply – Guide you through applying and saving notes (the system does not auto‑submit).
Interact using ChatGPT. When you run these commands, OpenCode forwards your prompt to ChatGPT. ChatGPT interprets the mode files and uses your CV and profile to generate tailored responses. The outputs (reports, PDFs, trackers) are stored in the same directories (reports/, output/, data/), so you can access them directly.
Customise and iterate. You can ask ChatGPT (through OpenCode) to modify archetypes, update your narrative or refine scoring criteria. The instructions in CLAUDE.md emphasise that any customisation should be stored in modes/_profile.md or config/profile.yml rather than editing system files.
Limitations and future work
No native OpenAI scripts yet. An issue opened in April 2026 suggests adding a native OpenAI integration with custom scripts (openai-eval.mjs, openai-pipeline.mjs) and npm tasks (openai:eval, openai:pipeline). As of now, this functionality isn’t merged into the main branch, so the recommended approach is to use OpenCode as described above.
Human confirmation required. The system intentionally avoids auto‑applying to jobs. It helps you evaluate, tailor and plan but leaves the final decision and submission to you. Be prepared to manually confirm or adjust applications.
Model variability. ChatGPT’s behaviour may differ from Claude’s evaluation logic. You may need to fine‑tune your prompts or adjust scoring weights in modes/_profile.md to align with your preferences.
Summary

The career‑ops repository is an advanced, open‑source job‑search pipeline that automates job evaluation, résumé tailoring, portal scanning and application tracking. While it was built on Claude, its modular design and open sourcing of mode files allow it to be used with other large language models. With a ChatGPT Plus or Pro subscription, you can run the pipeline through the OpenCode agent: connect your ChatGPT account, load the project, and issue /career‑ops slash commands. This approach leverages ChatGPT’s reasoning while benefiting from the structured workflow and scripts provided by career‑ops. The project is actively evolving; an open issue proposes native OpenAI scripts, so direct ChatGPT integration may become simpler in the future.

[discreteds]

How can we use it with codex

You can this to upgrade the plugin for other AI platforms: https://github.com/hiivmind/plugin-portability
You could do it for your own fork, or create a PR back into this repo.

[newway-anshul]

Thanks for the reply, I will check.