Six Claude skills that run your AI job hunt like an operating system. Decode the JD · research the company · tailor the resume · write outreach that gets replies · drill the interview · negotiate the offer.
Install · The skills · How they chain · Stay in the loop
Job hunting is a loop, not a lottery: decode what the role wants, prove you match it, reach a human, practice the room, negotiate the number. These skills make Claude run each step properly. They deliberately avoid frozen market-wide benchmarks: current compensation and hiring claims must be sourced when the skill runs.
What's inside
ai-job-hunt-os/
├── CLAUDE.md shared rules: your profile, honesty, human voice
├── README.md you are here
└── skills/
├── jd-decoder/
│ └── SKILL.md what the posting actually wants + apply/skip verdict
├── company-researcher/
│ └── SKILL.md the brief a well-connected friend would give you
├── resume-tailor/
│ └── SKILL.md your true story, told in the JD's language
├── cold-outreach-writer/
│ └── SKILL.md messages busy people actually answer
├── mock-interviewer/
│ └── SKILL.md realistic drills with honest scoring
└── salary-negotiator/
└── SKILL.md negotiation method + current-data sourcing rules
Install
Claude Code (30 seconds):
git clone https://github.com/landedjobs/ai-job-hunt-os.git
cp -r ai-job-hunt-os/skills/* ~/.claude/skills/
Then just talk to Claude: "here's a JD, should I apply?" and the right skill kicks in. Keep CLAUDE.md in your working folder (or paste it into your project's CLAUDE.md) so the shared rules apply.
Claude.ai (web/desktop): download this repo as a zip (green Code button), then upload the skill folders you want under Settings → Capabilities → Skills.
Any Agent Skills client: these are standard SKILL.md files per the open Agent Skills format; they work in any agent that supports it.
The skills
| Skill | What it does | Use it when |
|---|---|---|
jd-decoder | Separates real requirements from wishlist, translates the loaded phrases, gives an apply / stretch / skip verdict | Before every application |
company-researcher | Business, funding, AI-native or AI-washed, team intel, talking points, questions to ask | Before applying, again before the loop |
resume-tailor | Rewrites your resume for one specific JD without inventing anything, with a change log that teaches the pattern | Every application that passed the decoder |
cold-outreach-writer | Three message variants + follow-up plan, under 120 words, payload first | Reaching hiring managers and referrers |
mock-interviewer | Behavioral, technical AI/ML, or LLM system design sessions with real follow-ups and 1-5 scoring | The week before the loop |
salary-negotiator | Scripts for every negotiation moment + a method for sourcing current, comparable compensation | The screen question, and the day the offer lands |
How they chain
One target role, start to signed offer:
jd-decoder ──► company-researcher ──► resume-tailor ──► cold-outreach-writer ──► mock-interviewer ──► salary-negotiator
verdict the brief the resume the referral the reps the number
Each skill picks up the previous one's output, so you never paste the same JD twice. CLAUDE.md holds your profile so you never repeat your background.
Stay in the loop
These skills are the manual loop. landed is the agent that runs it for you daily: fresh AI-native roles matched to you, application answers drafted, interview prep queued.
Resume tailoring and cold outreach stay in this repo: they are open-source skills you run yourself in Claude. landed does not rewrite your resume or send outreach for you.
- ⭐ Star this repo: new skills get added (interview-question mining, offer comparison, visa-aware search are queued).
- 📬 Fresh roles: every open role landed tracks, by role family, refreshed daily with pay and the companies hiring.
- 🧰 More free tools: the other open-source repos landed maintains, from interview questions to portfolio projects.
License
MIT. Use it, fork it, share it with someone mid job hunt.