Artemis: an AI job-hunting partner

Artemis finds roles worth your time, tells you honestly how well you fit each one, and builds the
application: a tailored, ATS-safe resume, a cover letter, outreach to the right person, and
interview prep. She tracks every application and learns what works for you.

It runs in ChatGPT, Gemini and the Claude app with no install, and in Claude Code,
Codex CLI and Gemini CLI with the full toolset.

An illustrative run with the demo persona:

> artemis find
Apply: Senior Backend Engineer, Mollie (86/100). Payments + Go + ledger work match; gap: Kotlin depth.
Apply: Staff Engineer, Elastic (78/100). Remote Europe; gap: no search-engine experience, frame the Kafka pipeline.
Stretch: Backend Tech Lead, Wise (68/100). Hybrid London only: remote-restricted, check for an EOR.
Skip the other 23. Apply to Mollie and Elastic now.

What makes it different

• It never invents anything. Every resume line comes from your fact bank, a master CV
  where each fact has a source. If a job wants something you don't have, Artemis says so and
  suggests an honest framing. If a number would help, it asks you.
• Honest scoring. Each role gets a 100-point fit score (level, must-haves, location, domain,
  pay, company health, upside) and a verdict: apply, stretch or skip.
• Direct sources. It reads companies' own careers boards (Greenhouse, Lever, Ashby,
  SmartRecruiters, Workable, Workday), HN "Who is hiring?", Himalayas, VC portfolio boards, We Work
  Remotely and LinkedIn's public job search. It's not limited to aggregators.
• Draft only. It never applies, sends a message or logs in for you. You review everything
  and press send yourself.
• Your data stays yours. Your profile, pay and applications live in gitignored folders on your
  machine, or in your own chats and Projects.

Pick your setup

Start with your CV in any form: a PDF (LinkedIn's More → Save to PDF works), a Word document,
your LinkedIn profile link, or your whole LinkedIn profile copied and pasted. Then say artemis setup.

No install: ChatGPT, Gemini or the Claude app

Download the pack for your app from the latest release:

• ChatGPT (custom GPT): https://github.com/ciprianrosca/artemis/releases/latest/download/artemis-chatgpt.zip
• Gemini (Gem): https://github.com/ciprianrosca/artemis/releases/latest/download/artemis-gemini.zip
• Claude app (Project + Skills): https://github.com/ciprianrosca/artemis/releases/latest/download/artemis-claude.zip

Each zip has a HOW_TO.md with the steps (about 5 minutes):

• ChatGPT: create a custom GPT, paste INSTRUCTIONS.md, upload the three knowledge files, and
  turn on web search and code. See chat/platforms/chatgpt.md.
• Gemini: create a Gem, paste INSTRUCTIONS.md, and add the three knowledge files. See
  chat/platforms/gemini.md.
• Claude app: upload the six skill zips, create a Project, and paste PROJECT_INSTRUCTIONS.md.
  See chat/platforms/claude.md.

Then say artemis setup and attach or paste your CV. Say "try the demo" first if you'd
like to see it work on Sam Rivera, a fictional engineer.

A shared GPT or Gem doesn't keep your files between chats: keep the four files Artemis gives you
(master_cv.md, knowledge.json, contact.json, pipeline.md) and attach them to each new chat,
or add them to a Project once.

Full version: Claude Code, Codex CLI or Gemini CLI

git clone https://github.com/ciprianrosca/artemis.git
cd artemis
pip install -r requirements.txt      # renders DOCX/PDF and reads CV files
claude                               # or: codex, or: gemini
> artemis demo                       # try it as Sam, a fictional engineer
> artemis setup                      # then make it yours (about 10 minutes, starts from your CV)

Claude Code reads CLAUDE.md, Codex reads AGENTS.md and Gemini CLI reads GEMINI.md. All three
map the same commands to the same playbooks in skills/.

Commands

# Start
artemis demo                       # try it with Sam Rivera, a fictional engineer
artemis setup                      # onboarding: CV (PDF, Word, LinkedIn link or paste) → fact bank + settings
artemis profile                    # fill gaps in your fact bank

# Find
artemis find ["title"] [place]     # full sweep, deduped, scored, ranked
artemis scan                       # your watchlist's careers boards only (CLI)
artemis boards [hn|himalayas|getro|wwr]
artemis discover                   # scan ~1,000 tech companies for matches (CLI)
artemis watch add <company>        # add a company's careers board to your watchlist
artemis fit <url | pasted ad>      # scored fit, gaps, apply/stretch/skip

# Apply
artemis tailor <url>               # resume + cover letter → DOCX + PDF
artemis cover <url>                # cover letter only
artemis answers <url>              # drafts for application-form questions
artemis outreach <person|company>  # connection note, InMail, referral ask (text you send)
artemis linkedin people <company>  # hiring manager, recruiter, warm paths (public search)
artemis linkedin profile           # rewrite your headline, About and experience

# Track and decide
artemis pipeline | status | followups
artemis log <company> <stage>
artemis map                        # role families + target companies + pay reality check
artemis prep <company>             # company brief, likely questions, your stories mapped
artemis offer <details>            # gross → net, vs. your floor, negotiation levers
artemis recruiters                 # agencies and exec-search firms + outreach drafts

Plain language works too: "find me staff roles in Berlin", "is this a fit?", "prep me for Wise on Thursday".

How it's built

CLAUDE.md, AGENTS.md,    entry points for Claude Code, Codex and Gemini CLI
GEMINI.md
identity.md              who Artemis is, and the guardrails
skills/                  the playbooks: one SKILL.md per area, shared by all three surfaces
tools/                   standard-library Python: careers boards, job boards, LinkedIn public pages,
                         company discovery, CV reader, DOCX/PDF renderer, chat-pack builder
templates/               blank fact bank, settings, contact line, pipeline
examples/demo/           Sam Rivera, a fictional backend engineer, for trying things out
data/catalog.json        ~1,000 tech companies' careers-board slugs (unverified, PRs welcome)
data/countries/          offer maths and employment-law notes per country
chat/                    shared chat instructions + per-app setup (ChatGPT, Gemini, Claude)
profile/ applications/ memory/   YOUR data, gitignored

Every personal setting (target titles, where you can work, remote rules) lives in
profile/knowledge.json, so the tools contain nothing specific to any one person.

Good to know

• LinkedIn: Artemis reads public, logged-out pages only, a few requests at a time, and stops
  when LinkedIn rate-limits. It's for one person's search, not bulk collection. Using it is your
  call under LinkedIn's terms; everything else works without it.
• Pay, tax and law: offer maths comes with sources and dates, but it isn't financial or legal
  advice. Have a professional check anything that decides a contract.
• Your data: profile/, applications/ and memory/ are gitignored. If you fork this repo
  publicly, they stay on your machine.

Contributing

Anyone can propose a change; every pull request is reviewed and approved by the maintainer.
See CONTRIBUTING.md. Most useful:

• new country packs in data/countries/ (see the README there)
• more careers-board slugs in data/catalog.json
• new job sources in tools/, which must use public APIs or feeds and the standard library only
• improvements to the playbooks in skills/. Run python tools/build_chat_pack.py to rebuild the
  ChatGPT, Gemini and Claude packs from them
• fixes when a job board changes its format

License

MIT. Built by Ciprian Rosca with Claude.