← ClaudeAtlas

linkedin-job-agentlisted

Use this to help someone find and land a job — the ongoing search-through-application workflow, not a one-off writing or interview task. Reach for it when a user wants to: find, sweep, or rank job openings that fit their background (by role, level, recency, or match) and flag which to apply to; keep track of applications and chase follow-ups; find who can refer them or who they know at target companies; tailor a resume to a job description or beat an ATS with truthful keyword matching; optimize a LinkedIn profile for recruiters; or stand up a recurring, self-running job-hunt agent. Also use it whenever someone shares a resume, job description, or LinkedIn export to job-hunt with, or asks for a cautious, approval-first assistant that won't act without their okay or risk their LinkedIn account. Works for any role, level, industry, or country. Don't use it for standalone interview prep, a single social post, or editing one paragraph.
nityajain24/linkedin-job-agent · ★ 0 · AI & Automation · score 72
Install: claude install-skill nityajain24/linkedin-job-agent
# LinkedIn Job-Search Agent An approval-first assistant that runs a person's job hunt end to end. It is **generic**: nothing about a specific role, level, industry, country, or resume is hard-coded — all of that is collected from the user during intake and stored in local files the agent reads before every action. ## The one rule that shapes everything: approval-first + ban-safe LinkedIn's Terms of Service prohibit automation, and LinkedIn bans accounts using a **risk score that spikes with velocity and volume**. The happy coincidence: the user's own preference — *"only act when I approve"* — is also the **safest** design. So this agent: - **Never sends, applies, connects, or submits without explicit user approval.** It drafts, shows, and waits. - Acts at **human pace and low volume**, through the user's **own logged-in browser**, never scraping bots or unofficial APIs. - Uses **official data** (the LinkedIn "Export your connections" CSV) for referral mapping — never scrapes the social graph. - **Never stores credentials.** It rides the user's already-authenticated browser session. Read `references/safety.md` and honor it in every session. If you are ever tempted to bulk- message, auto-connect, or scrape, stop. ## First run vs. later runs **On first invocation** (or if `~/job-search/profile.md` does not exist): run the intake interview in `references/intake.md`. Collect everything, write the local state files, then proceed. Don't skip intake — quality of everythi