search

name	search
description	Search Australian job listings (Adzuna + Workforce Australia) and rank them against the user's profile. Use when the user wants to find jobs, e.g. "/jobfinder:search devops engineer Brisbane" or "find me remote data roles".

jobfinder: search

Find Australian jobs and rank them against the user's grounded profile. $ARGUMENTS is the free-text request.

Steps

1. Load grounding. If ~/.jobfinder/profile/profile.md exists, Read it into context now. If it doesn't, tell the user to run /jobfinder:profile first — you can still search, but ranking will be generic without it.

2. Build the query. Translate the request (plus profile defaults) into flags. Interpret bare city names as Australian (e.g. "Sydney" → --where Sydney). Geolock: never search another country or substitute a non-Australian location. Then run:

   uv run "${CLAUDE_PLUGIN_ROOT}/scripts/search.py" --what "<keywords>" [--where "<AU location>"] [--salary-min <AUD>] [--max-days-old <N>]
   

   The script returns JSON: {query, new_count, total_found, listings:[...]}. Each listing has id, title, company, location, salary_min/max (AUD), url, source, and a description snippet. Newly-seen jobs are stored automatically. If listings is empty, broaden the query (drop the location, widen --max-days-old, loosen keywords) and try again.

3. Score each listing 0–100 against the profile using this rubric:

   • Skills overlap — 40: how many of the user's core skills the role needs.
   • Seniority fit — 20: role level vs the user's seniority.
   • Location / remote fit — 15: matches preferred AU locations or remote.
   • Salary — 15: listing salary vs the user's AUD expectation (unknown salary = neutral, don't penalize hard).
   • Recency — 10: more recent postings score higher.
   • Red-flag penalty: subtract for hard mismatches (work rights the user lacks, an explicit red-line, wrong field).

4. Persist the scores so the tracker and alerts can reuse them:

   uv run "${CLAUDE_PLUGIN_ROOT}/scripts/tracker.py" score <id> <0-100> "<one-line reason>"
   

5. Present a ranked table (highest score first): score, title, company, location, AUD salary, source, id. Add a one-line "why it fits / watch-out" for the top handful. Be honest about weak matches — don't inflate scores.

6. Offer next actions: save a job (/jobfinder:track save <id>), tailor an application (/jobfinder:tailor <id>), or save this as a recurring alert (/jobfinder:alerts save <name> <query>).