OPERATOR BRIEFING // ABOUT THIS SYSTEM

What you're looking at, Abdurrahman.

A 60-second read. Skim it once, you'll know how the whole thing works.

The premise

This is a personal job-search co-pilot. It does the boring work (scanning postings, ranking them, drafting tailored materials) so you can spend your time on the parts only you can do (deciding which to chase, writing real things about your actual experience, hitting submit).

Think of it as a war room with one operator: you. The dashboard is the situation board. The scripts run reconnaissance. Claude Code is the planner sitting next to you. You stay in the cockpit at all times — nothing gets sent without your finger on the button.

The daily flow

Morning
In Claude Code, run /find-jobs. It scrapes LinkedIn / Indeed / Google / ZipRecruiter plus the ATS boards for companies you flagged. It hard-filters by your dealbreakers (salary floor, location, target titles, hard-no industries), then triages survivors into CRITICAL / ENGAGE / RECON / ARCHIVED using the rubric in CLAUDE.md. This page (the queue) refreshes.
Pick something
Scroll the queue. Read the one-line reason. Decide what you actually want to chase today. CRITICAL = sweet-spot fit. ENGAGE = strong fit, worth applying. RECON = closer read needed. ARCHIVED = skipped, kept for the record.
Prep one
Click ⬢ JACK IN on a card. It copies /prep <slug> to your clipboard. Paste it back into Claude Code. Claude generates a tailored resume, drafts a cover letter (using the humanizer chain to strip AI tells), and builds a launchpad page with copy-buttons next to every form field. Estimated time: 2–5 minutes per application instead of 30–60.
Submit yourself
Open the launchpad. Open the apply page in another window. Alt-tab between them. Click-copy your way through. You always click submit. Never the system.
Track
Run /sync-mail to pull recent Gmail and update timelines on active applications (rejection / interview / radio silence).

What this system does NOT do

It never submits an application. Drafts only.
It never sends an email or message on your behalf. All outreach is drafted; you send.
It never touches credentials, banking info, SSN, ID documents, or work-auth paperwork.
It never edits your resume or preferences without you asking. Those files are the source of truth.
It does not invent experience. If a posting asks for something you don't have, it flags the gap in the fit analysis. Stretching the truth on an application is the failure mode we're most afraid of.

The files that matter

If you only ever edit four files, edit these:

profile/resume.md — your canonical career history. Tailored versions get generated per-application; this stays untouched.
profile/preferences.md — salary floor, locations, target titles, hard-no companies, ethos signals. The triage rubric reads this every time it runs. If the queue is full of garbage, fix this file first.
profile/stories.md — concrete anecdotes from your actual work. The single most important file for cover-letter quality. Right now it's a template. Until you fill it, cover letters will sound generic.
profile/voice-samples.md — 1–2 pages of your real writing in different registers (a Discord message, an email, a casual DM). The humanizer chain reads this to calibrate. Same situation as stories.md — needs you.

Tier glossary

CRITICAL — sweet-spot fit (junior dev at a target / ethos company). Apply first.
ENGAGE — strong fit with one or two gaps that can be framed. Apply.
RECON — meaningful gaps. Read the JD before deciding.
ARCHIVED — skipped. Kept on the dashboard at low opacity so you can see what was filtered out and adjust preferences if needed.

Any tier can carry chips: TARGET CO. (you flagged this company as always-surface), ETHOS+ (matches your "place that treats its people well" exemplars), +remote-first / −fast-paced (signals scanned from the JD body that nudge the score).

Quick command reference

From inside Claude Code, in this project directory:

/find-jobs       refresh discovery → filter → triage → dashboard
/triage          re-score the existing queue (use after editing preferences)
/prep <slug>     build full application bundle for one job
/cover <slug>    regenerate just the cover letter for an existing prep
/sync-mail       pull recent Gmail, classify, update timelines

The honest part

Discovery and triage work right now. They'll surface 30–80 postings per run, and the rubric is consistent enough that the top of the queue is usually worth reading.

Cover letters are a different story. Until profile/stories.md and profile/voice-samples.md have your real input, the humanizer chain has nothing distinctive to work with — drafts will sound like AI-edited resume copy. Plan ~30 minutes of voice memos (situation → what YOU did → what was hard → outcome → what stuck) and ~10 minutes of pasting your own real writing. That single hour is the highest-leverage thing left to do.

◄ BACK TO FEED