job-hunter

agent
Guvenlik Denetimi
Basarisiz
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Basarisiz
  • rm -rf — Recursive force deletion command in install.sh
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Skills for Claude Code, Codex and Cursor: paste a job link, get your CV scored and rewritten for it, the people to reach out to, and the message and cover letter to send. No API key.

README.md

jobhunt

Eleven skills for the coding agent you already use. Keep everything you have
ever done in one place; paste a job link and get a CV built for it, the people
worth messaging, and the message to send them.

Install → · pick your
agent and copy one command.


What it takes off your plate

An application used to cost you an evening: tailoring the CV, working out who
to contact, writing the message. Every time, for every job.

Your work lives in one profile. Every role, project and side thing across
every field you have worked in. You write it once, and each application draws
from it — you never retype your history.

A CV per job, without the evening. It picks which of your work answers this
posting and leads with it. Same facts, different order:

The one CV you send everywhere        Built for this posting
· Responsible for various tasks       ● Cut ETL runtime 35% (dbt models)
· Worked on internal tooling          ● Own the ingestion pipelines
· Involved in cross-team projects     ● Built the dashboards 40 people use
● Cut ETL runtime 35% (dbt models)    · Responsible for various tasks

1 of 3  in the first screenful        3 of 3  in the first screenful

Same six lines from your own profile. Your strongest three were below the fold,
and most readers never reach them.

The message, already written. It works out who at the company is worth
contacting — alumni first, because that is what actually gets replies — builds
the searches that find them, and drafts something specific enough to answer.
You press send.

Coming soon

One-click apply on the boards that allow it, jobs found for you, and
application tracking. Not built yet — today it prepares the application and
hands it to you; the sending is still yours.

Install

Pick your agent on the install page,
or:

# any agent, macOS or Linux
curl -fsSL https://raw.githubusercontent.com/mohamedaminehamdi/job-hunter/main/install.sh | sh

# Cursor, Cline and Windsurf read skills per project, not per user
curl -fsSL .../install.sh | sh -s -- --to .cursor

Inside Claude Code or Codex you can install it as a plugin instead:

/plugin marketplace add mohamedaminehamdi/job-hunter
/plugin install jobhunt@jobhunt

Or download a folder from the install page and drop it in. No terminal needed.

What it needs: Python 3.9 or newer, which macOS and every Linux already
has, and Chrome, Chromium, Edge or Brave for reading job pages and making PDFs.
No API key — your agent is the model, so whatever you already pay for covers
it.

Optional: a TeX engine. If you have one — tectonic is a single binary —
CVs and letters are set in LaTeX instead, which is what most people expect a CV
to look like and is several times faster. Nothing needs it: with no engine the
browser renders them exactly as before, and the tool tells you which it used.

Using it

Put your CV somewhere, open your agent in a folder you want to work in, and
say what you want:

prepare an application for https://boards.greenhouse.io/acme/jobs/42

The first run reads your CV, writes jobhunt/profile.yaml, and stops and
asks you to check it
. A model just read your career; you approve it once, and
everything after is built on facts you have agreed to.

Then you get a folder per job:

jobhunt/runs/2026-09-24-acme-senior-data-engineer/
├── fit-before.md     how your CV answered this job as it stood
├── cv.pdf  cv.md     the tailored CV
├── letter.pdf .md    the cover letter
├── fit-after.md      what the tailoring actually bought
├── critique.md       what is still weak in both
├── outreach.md       who to message, the searches to run, what to say
└── job.yaml          the posting, as read

The eleven

Each works on its own. Install just the review to score your CV, or just the
fit score to decide whether a job is worth an evening.

jobhunt the whole thing, in order
jobhunt-profile your CV → one YAML file everything else reads
jobhunt-review score the CV on its own, with no job description
jobhunt-posting a job URL → structured posting, in your own browser
jobhunt-fit how well you match, before and after
jobhunt-tailor a CV for this job, out of what you have already done
jobhunt-letter three or four paragraphs worth reading
jobhunt-pdf the files to attach
jobhunt-answer form questions, including how to write an honest no
jobhunt-outreach who to message, and what to say
jobhunt-critique what's still wrong, before you send it

What it does not do today

  • Apply to anything. It produces the documents; you send them. On the
    roadmap for the boards that allow it — not built yet.
  • Scrape LinkedIn. LinkedIn walls and throttles automated access and the
    risk lands on your account. It builds the searches; you run them, and that
    is not changing.
  • Search for jobs. You bring the link, for now.
  • Track your applications, beyond one line per run in runs/log.md. It is
    markdown; type what happened next into it.

When it goes wrong

What you see What it means
No Chrome, Chromium, Edge or Brave found Install any of them, or set JOBHUNT_BROWSER. You still get markdown without one
Only 300 characters came back The posting is behind a login wall. Paste the description — that path is fully supported
No profile at ... Put a CV somewhere and ask for a profile first
is YAML this reader does not do Hand-edited profile.yaml using a {a: b} map. Use block style
The fit score says 0 of 0 The posting stated no requirements, so it fell back to keywords. Coarser, and it says so

Contributing

See CONTRIBUTING.md, and AGENTS.md for how the
repo is laid out. pytest — 444 tests, on Python 3.9 and up.

License

MIT — see LICENSE.

Yorumlar (0)

Sonuc bulunamadi