JobMentis: bring in the job search you built in ChatGPT or Claude

The prompt, free to copy

It asks your assistant for your own words and your own numbers only, and to leave out anything an employer may not ask about.

Copy everything below the line into the assistant where your job-search work already lives (a ChatGPT chat or project, a Claude project, a Gemini conversation, or any assistant that can read your documents). Run it there. It will produce one file. Paste that file back into JobMentis. If your system lives in a document rather than a chat, attach the document first, then run the prompt.

You are helping me move my job-search system into JobMentis, a career preparation platform. Read everything you have about me in this conversation, this project, and any files I have attached: my CV, my notes, my target roles, my stories, my preferences, our past conversations about my search.

Produce ONE reply, structured in two parts.

**Part 1: a short summary for me to read.** Five to ten lines: who I am professionally, what I am looking for, and what you found. Write it in my language.

**Part 2: a single fenced code block** exactly as specified below, so a machine can read it. Use the tag `jobmentis-import` on the fence.

```jobmentis-import
{
  "schemaVersion": 1,
  "language": "<the ISO code of the language I write in, e.g. en, fr, it, es, el, de>",
  "profile": {
    "headline": "<how I describe myself professionally, in my own words, or null>",
    "targetRoles": ["<role titles I am targeting>"],
    "targetCompanies": ["<companies or types of company I am targeting>"],
    "locations": ["<cities or countries I will work in>"],
    "workModes": ["<onsite | hybrid | remote, only those I said>"],
    "languages": ["<languages I work in, with level if I stated one>"]
  },
  "compensation": {
    "currency": "<e.g. EUR>",
    "period": "year",
    "min": <the lowest number I said I would accept, or null>,
    "ideal": <the number I said I am aiming for, or null>,
    "current": <what I earn now, only if I stated it, else null>
  },
  "workHistory": [
    {
      "employer": "<company>",
      "title": "<my title>",
      "start": "<YYYY-MM>",
      "end": "<YYYY-MM or null if current>",
      "location": "<city or country, or null>",
      "summary": "<one or two lines on the scope of the role, or null>",
      "achievements": ["<one achievement per line, in my words, with the numbers I actually gave>"]
    }
  ],
  "stories": [
    {
      "title": "<short name for the story>",
      "framework": "STAR",
      "situation": "<...>",
      "task": "<...>",
      "action": "<...>",
      "result": "<...>",
      "tags": ["<themes: leadership, conflict, turnaround, ...>"]
    }
  ],
  "skills": ["<skills I claimed>"],
  "certifications": ["<certifications I hold>"],
  "education": ["<degree, institution, year>"],
  "preferences": {
    "dealbreakers": ["<things I said I will not accept>"],
    "musts": ["<things a role must have for me>"],
    "workingStyleNotes": ["<how I work best, in my words: autonomy, pace, structure, management style I need>"]
  },
  "searchRules": [
    { "rule": "<a rule I follow, e.g. do not apply below X, avoid companies with Y>", "reason": "<why, if I said>" }
  ],
  "contacts": [
    { "name": "<person>", "company": "<company or null>", "role": "<role or null>", "relationship": "<how I know them>", "linkedinUrl": "<url or null>" }
  ],
  "notes": ["<anything important that did not fit above>"]
}
```

Rules, and they matter more than completeness:

1. **Use only what is actually in our conversation or my files.** Never guess, never round up, never smooth out a gap. If you do not know something, use `null` or leave the array empty. An empty field is correct; an invented one is a problem I will only discover in an interview.
2. **Keep my wording** for achievements, stories and preferences. Do not rewrite them into corporate language.
3. **Numbers must be mine.** Only include a figure if I stated it. Never estimate a result, a budget, a team size or a salary.
4. **Do not include** anything about my health, disability, religion, political views, trade union membership, ethnic origin, sexual orientation, sex life, date of birth, age, marital status or children, even if it appears in my files. Skip it silently.
5. **Do not include** my address, phone number or national identifiers. My name and email are fine.
6. **Cap the volume**: at most 15 roles, 20 stories, 40 skills, 30 contacts. If there are more, keep the most recent and the most relevant to my target roles.
7. **One code block, valid JSON, no comments, no trailing commas.** If a value contains a quote, escape it.

When you are done, tell me in one line how many roles, stories and contacts you exported, and what you deliberately left out.

What never comes in

• Health, disability, religion, political or union details, orientation, family, age and nationality: left out before you see the review, and only counted.
• Your current salary: shown to you, never stored.
• Anything you untick.

Questions people ask

Prefer to keep running your search inside Claude? See the open-source Claude skills