codex-job-finder

Codex skill-first workflow for job research automation.

Codex runs the workflow. TypeScript CLIs do the deterministic work. SQLite keeps the facts. Markdown makes the result reviewable.

Demo

job_finder_demo_30s.mov

Install

git clone https://github.com/SongMarco/codex-job-finder.git
cd codex-job-finder
pnpm install
pnpm test
pnpm typecheck

Use In Codex

Run the skills in this order:

$search-jobs
Find Node-family backend JD candidates.

$jobplanet-company-profiles
Fill missing company profile enrichment for the collected companies, using exact matches only, and report status counts before and after.

$query-jobs
Query and show me top 20 jobs.

$db-markdown-report
Create a Markdown report from the current SQLite database with a short summary first and the full result table below.

What Each Skill Does

Skill	Role
$search-jobs	Collects and normalizes JD candidates into job_listings
$jobplanet-company-profiles	Enriches companies in company_profiles
$query-jobs	Shows JD candidates with company score fields and detail URLs
$db-markdown-report	Exports DB summaries or query results as Markdown

Skills live in:

.codex/skills/

Default local outputs:

data/headhunter.db
results/search-jobs/
results/jobplanet-company-profiles/

data/ and results/ are gitignored. This public repo includes only synthetic fixtures and sample reports.

Safety

• Prefer deterministic CLI/module execution over manual browser scraping.
• Do not use cookies, browser profiles, local/session storage, vault data, or secrets.
• Do not bypass CAPTCHA, authwalls, MFA, OAuth/risk prompts, Cloudflare, login walls, or rate limits.
• Treat blocked access as blocked.
• Keep real databases and real reports out of the public repo.

More

• Sample report
• Architecture
• Case study