Home › Projects › RESUME.md

Anbeeld's RESUME.md — AI Resume/CV Rules

GitHub

AI ruleset to build or improve CV that interviews you step by step and works for any market

A ruleset that governs the full pipeline from zero candidate information to finished CV, across markets, languages, and model strengths. Not just bullet-writing tips: fact locking, evidence forcing, positioning, self-audit, market conventions, and file export are all in scope. Works with any AI that takes instructions and adapts to whatever input you give it, from a blank screen to a decade of career history.

The rules combine career-consulting methodology, ATS extraction research, hiring-side scanning behavior, and direct observation of LLM defaults: underselling candidates, narrowing to the latest niche, skipping evidence, leaking internal process language, and writing generic duty-list bullets.

What it does

• Positions around the full career record and target preference, not the latest job. Seniority calibrates to ownership and production responsibility, not to adjectives or whatever title the company happened to use.
• Forces evidence before a role closes. If metrics are missing, the AI asks down a six-rung ladder from direct outcomes to qualitative anchors instead of accepting "I did the work."
• CVs that work for both the parser and the person. ATS extraction, recruiter scanning, and AI matching all read the same text, without keyword stuffing or keyword starvation.
• Every draft passes a self-audit before the human sees it: factual accuracy, positioning range, evidence strength, ATS-parsing safety, render integrity, and a red-flag search that strips weasel verbs, generic phrases, and leaked internal process language.
• Invented claims are refused. Ask for something false and the ruleset states the risk once, then offers the strongest truthful alternative.
• Confirmed facts stay locked. The AI does not re-ask them, does not draft before evidence and positioning are in place, and does not narrow the candidate to the latest niche before the full history is known.
• Market conventions shift per region across ten defaults covering major hiring markets, so one ruleset handles different expectations without averaging them into a compromise.

How to use it

Give the ruleset to ChatGPT, Claude, Gemini, or any other AI (preferably with thinking turned on), tell it to follow every rule strictly, and ask to build or improve your CV. At ~7700 words, the instructions are specific: fact locking, evidence forcing, positioning discipline, market-aware output, and a built-in self-audit.

• Starting from nothing? It asks one question at a time so you never have to dump your whole career at once.
• Have an existing CV, LinkedIn profile, portfolio, or GitHub page? It reads what is there, locks the facts it finds, and asks only for what is missing.
• Tailoring to a job posting, translating for a different market, or just asking a single question? It chooses the right scope instead of forcing the full pipeline every time.

After the first draft, the self-audit and red-flag search run before you see anything. From there, you can correct details, add missing facts, or ask to shift the positioning, and each revision gets sharper without losing what was already right.