Score a LinkedIn job posting against your own profile in one click
Idea + starter prompt, not run by us yet
For: Job seekers who read dozens of postings and waste time on poor fits
A job seeker kept his career profile and a weighted scoring prompt, then turned them into a Claude in Chrome shortcut. On any open LinkedIn posting he picks the shortcut and gets a fit report with a match percentage and a verdict. It caught that one role was on-site, which clashed with his remote-only profile.
Steps
- Write your career profile as a plain file: skills, years, salary floor, location rules and deal-breakers.
- Host the profile somewhere the extension can fetch, such as a private GitHub Gist, so you edit it in one place.
- Create a Claude in Chrome shortcut that holds your scoring rules with a weight for each factor.
- Open a LinkedIn posting, type / and pick the shortcut, then read the report.
- If you also keep the rules as a Claude Skill, remember the shortcut is a separate copy. Change both when your criteria change.
Starter prompt
First, a trial. Score a short made-up job ad that I paste below, so I can check your report format before you touch a real page. This job only reads and drafts. It changes nothing on LinkedIn and writes only in this chat. After my OK, read the one LinkedIn posting open in this tab. Do not click Apply, save the job, follow anyone or message anyone. If a button might do any of that, stop and ask for my yes first. Showing me a plan is not permission to click. My profile is at [link or pasted text]. Weights: [for example skills 40%, location and remote rules 30%, pay 20%, seniority 10%]. Score the posting from 0 to 100%. List what matches, what does not, and anything the posting hides or leaves vague. Check location and remote rules first and say clearly if the posting conflicts with my profile. End with one verdict: Apply, Maybe or Skip, with two sentences of reasoning. Never ask me for a password.
Original post: automato.substack.com · found via agentusecases.ai. Written up in our own words. We haven't run it, so test it on made-up data first.