Open the spreadsheet you use to track your job hunt. Be honest about what it actually does for you. Most of them are a list of companies you applied to, a column for status that you last updated two weeks ago, and a creeping sense that you have lost track of the whole thing.
That spreadsheet is not a system. It is a guilt file. It records what you already did and shames you for not doing more. It does not judge whether a role is worth your time before you apply. It does not remember the roles you rejected, so you risk applying again. It does not know who you are.
Two markdown files do all of that better. Not a spreadsheet. Not an app. Two text files.
File one: cv.md, who you are
The first file is cv.md. This is the single source of truth for who you are professionally. Your roles, the work you did, the real metrics behind it, your skills grouped the way you actually think about them, the roles you want next and the ones you are moving away from.
The rule that makes this file powerful is strict. Every score, every gap, every verdict the system produces reads this file. You never hardcode a fact from it anywhere else. Not in the evaluator, not in a prompt, not in a cover letter template. The CV is the one place your story lives, and everything else reads it fresh.
You do not put in any filler, and you do not get a filler match.
That last line matters more than it sounds. The reason AI-generated CVs feel generic and slightly dishonest is that the input was generic. The model was asked to invent a person. When the input is your real experience, with your real numbers, the output is anchored to the truth. The match it produces is real because the source is real.
File two: CLAUDE.md, how you decide
The second file is CLAUDE.md. This is how you decide. It holds three things.
Your scoring rubric. Every role gets a score from one to five across five dimensions: match with your CV, fit with the roles you actually want, the compensation signal, culture and red flags, and posting legitimacy, meaning is this a real opening or a ghost job.
Your dealbreakers. The conditions that make a role a non-starter no matter how good the rest looks. No B2B-only contracts if you are a contractor. Remote only. Will not relocate. A salary floor. Whatever yours are. If a dealbreaker appears in a posting, the call is SKIP, full stop, regardless of score.
The guardrail. The AI never auto-submits anything. It evaluates and recommends. You decide and act. You always look at the draft before anything goes out.
Why two files, not one
Splitting identity from judgment is the move. If you mash them together, every time your CV changes you risk rewriting your decision rules, and every time your dealbreakers change you risk touching your history. Keep them apart and each one stays clean.
There is a deeper reason. The two files map to the two bottlenecks that wreck a job hunt. Judgment is which roles are worth you. Memory is never forgetting and never applying twice. cv.md carries your judgment in a form the system can reason about. CLAUDE.md encodes the rules that apply that judgment the same way every time. Together they handle both bottlenecks, and the tracker, which logs every verdict, handles the memory.
The same pattern, pointed at a new problem
If you have been building along, you have seen these files before. CLAUDE.md is the file that made your code standards permanent instead of pasted. cv.md is the same idea applied to your professional identity. The discipline is identical: put the thing you keep re-explaining into a file once, and let the system carry it on every run.
The job hunt is just the latest place that discipline pays off. Your identity stops being something you retype into a hundred application forms and starts being a file the system reads. Your judgment stops being a feeling you apply inconsistently and starts being a rubric that runs the same way every time.
Start tonight
Open a text editor. Start cv.md with your real experience and your real numbers. Start CLAUDE.md with your rubric, your dealbreakers, and the guardrail. Two files. That is the whole foundation. Everything else, the evaluator, the tracker, the tailored CV variants, sits on top of these two.
The spreadsheet was tracking what you already did. These two files decide what you do next. That is the difference between a guilt file and a system.