Daily Technical Writing Job Search

A Python script, written with Claude's help, that checks about ten job sources every day — company career pages, aggregators, and niche boards — filters for technical-writing roles you're actually eligible for (remote worldwide, or hybrid within a commute you'll accept), and emails you only the ones you haven't seen before.

It does not replace direct outreach or networking. In a niche field like technical writing, this is a safety net that makes sure you never miss a posting — it won't manufacture opportunities that don't exist yet.

Provided as is. It worked for its author, but job-board APIs change without notice, and I can't promise support or results. Use it, adapt it, and share it.

• Windows, Mac, or Linux with Python 3.8 or newer. Check with python --version (Windows) or python3 --version (Mac/Linux).
• The requests library. Install it once: pip install requests (Windows) or pip3 install requests (Mac/Linux). Everything else the script uses is in Python's standard library.
• A free Adzuna account (step 1 below).
• An email account you can send FROM — Gmail, Outlook, or similar (step 2 below).

Adzuna is the main job-search source. Sign up at https://developer.adzuna.com/signup, then create an "Application" in their dashboard. You'll get an App ID and an App Key — copy both somewhere safe for step 2. The free tier allows 250 calls/day; this script uses roughly 20–60 per run, so you won't come close to the limit.

1. In the same folder as job_search.py, copy secrets.env.example to a new file named exactly secrets.env (remove .example — not secrets.env.txt, which is a common accidental result of renaming in Windows Explorer; run dir secrets* afterward to check the exact name).
2. Open secrets.env in a text editor and fill in your real Adzuna App ID/Key and your email details. Follow the comments in that file — Gmail needs an App Password, not your normal password; the file explains how to generate one.
3. Never share this file or paste it anywhere public. It's the one file in this whole setup that's genuinely private.

Open job_search.py in a text editor (VS Code, Notepad++, anything) and find the section headed:

2A. PERSONALIZE  -  edit every value in this block for your own situation.

It's near the top of the file. Every value in that block is something specific to you:

• COUNTRY_CODE — which country's job market Adzuna searches. "ca" for Canada, "us", "gb", "au", "de", etc.
• OK_REGIONS — words that mean "I'm eligible" when they appear in a remote listing's location. Include your own country, common alternate names for it, and the usual "open to anyone" phrases.
• US_ONLY — phrases that mean "not eligible." Add any other country/region that keeps showing up that you can't work from.
• HOME_LABEL / HOME_COUNTRY_TERMS — how your own country is named in the emailed report, and the specific words (lowercase) that name your country/province/state. These feed into some of the eligibility logic beyond the simple lists above.
• HYBRID_CITIES — cities within a commute you'd actually accept, for hybrid roles. Anywhere else hybrid is found, it's still shown, just flagged with its city so you can judge it yourself.
• LOCAL_WHERE / LOCAL_DISTANCE_KM — Adzuna's local-radius search pass. Set LOCAL_WHERE = "" to turn this pass off entirely.
• MIN_SALARY_THRESHOLD — the lowest annual pay you'd consider.
• EXCLUDED_COMPANIES — any companies/staffing agencies you never want to see again.

Everything after that block (title keywords, scam phrases, the list of companies under COMPANY_BOARDS) is a shared starting point that should work reasonably well as-is — but feel free to tune it too, especially COMPANY_BOARDS. Adding more companies there is the single best way to get more results — it's a thin market, and the company list determines how much of it you actually see. Instructions for adding a company are right above that list in the file.

Open a terminal in the script's folder and run:

python job_search.py --dry-run

(On Mac/Linux, use python3 instead of python.)

This fetches and filters everything but sends no email and doesn't record anything as "seen" — safe to run as many times as you like. (If there are new jobs, it still saves a jobs_report_*.html file you can open.) Watch the console output: it prints one line per source, a funnel showing how many jobs survive each filtering step, and (if anything looks off) exactly which jobs got dropped and why. If a source says "not found" or shows an error, that's worth reading — it usually means a company slug needs fixing.

Once that looks right, drop --dry-run to send yourself a real email and start recording history:

python job_search.py

Add --force to any run to treat every matching job as "new" — useful the very first time, to see everything currently open rather than just future changes.

1. Find your real Python path. In PowerShell: where.exe python (not the plain where alias — it does something different). Use the path under Programs\Python\..., not the one under WindowsApps — that one is a Microsoft Store stub, not a real interpreter, and won't run reliably from a scheduled task.
2. Open Task Scheduler → Create Basic Task. Name it, set the trigger to Daily at a time of your choosing (morning works well, so boards have updated overnight).
3. Action: Start a program.
   • Program/script: the full python.exe path from step 1.
   • Add arguments: the full path to the script in quotes, e.g. "C:\Users\you\jobsearch\job_search.py"
   • Start in: the folder itself, e.g. C:\Users\you\jobsearch (no quotes, no trailing backslash). This is optional: the script looks for secrets.env and its history file next to itself, wherever it's launched from.
4. Finish, then right-click the task → Properties.
   • If you sign in with a PIN rather than a full password, leave "Run only when user is logged on" selected (the default) — no password needed. This runs fine from a locked screen, just not after a full shutdown.
   • On the Conditions tab: uncheck "Start the task only if the computer is on AC power" (otherwise it silently skips on battery), and check "Wake the computer to run this task" if you want it to fire even while asleep.
5. Right-click → Run once to test, then check your email and the folder for a new jobs_report_*.html file.

crontab -e

Add a line like this (adjust the path and time; 0 7 means 7:00 AM daily):

0 7 * * * cd /path/to/jobsearch && /usr/bin/python3 job_search.py >> run.log 2>&1

The >> run.log part saves the console output somewhere you can check later, since cron won't show it to you directly.

• "No such file or directory" — the filename you typed doesn't match what's actually on disk. Run dir *.py (Windows) or ls *.py (Mac/Linux) to see the real filename.
• Email never arrives — check spam, and re-read the console output; it prints "Report emailed to " or "Email FAILED: " every run.
• A source shows "0 fetched" with errors — not necessarily a problem; read the specific error. A timeout usually clears up on the next run.
• "Boards not found: delete these from COMPANY_BOARDS" at the end of a run — those company names are wrong (HTTP 404). Delete them from COMPANY_BOARDS, or fix the slug by checking the company's job-posting URL.
• Everything shows as "new" again after it worked yesterday — this means job_history.db wasn't updated, almost always because email failed on the previous run (jobs are only marked "seen" after a successful send). Fix the email problem and it'll catch up automatically.
• Environment variables set with setx/export don't seem to work — use secrets.env instead; it's more reliable, since variables set in one terminal window often aren't visible in a different one. (The script reads ordinary environment variables too, which is handy in containers or CI. If a variable is set in both places, the value in secrets.env wins.)

• The market is thin. Out of several thousand jobs fetched per run, expect single digits to survive every filter on most days. That's not a bug — it's genuinely how small the pool of remote/eligible technical writing roles is at any given moment.
• Not every company runs on a public-API platform. Some use fully custom career portals with no API to connect to. If a specific company you care about isn't showing results, check their careers page by hand — the script may simply have no way to see it.
• A few sources are newer and less proven than the core ones (Greenhouse, Lever, Ashby, Adzuna): Teamtailor and SmartRecruiters were added based on their documented public APIs but haven't been tested against as much real-world data. Keep an eye on their lines in the console output for a while.
• Some postings are caught by their description, not their tag. A board may label a job "Anywhere" while the text says "Remote (United States)". The script checks descriptions for common US-only wording and drops those jobs, but it can't catch every phrasing, so always read the posting before applying. That check looks for US wording only; if you're outside the US, it won't catch "UK only", "EU only" and similar.
• Free APIs ask to be used considerately. A few of these sources are small, free services asking for reasonable call volumes, not relentless polling. Running this more than once a day isn't necessary and risks getting rate-limited.

If you fork or copy this to your own repository, never commit secrets.env, job_history.db, or the jobs_report_*.html files. The included .gitignore excludes them. Also replace the example values in the personalization block with your own, or leave them as placeholders.

MIT. See LICENSE.