Build resumes from a long memory.

The verbose resume idea

1. Maintain a private verboseResume.md with more detail than any final resume should include.
2. Use stable headings: ## Experience, ### Company — Title, #### Notable work, ##### Project.
3. Capture projects, tools, incidents, team leadership, metrics, context, and accomplishments while they are fresh.
4. Update it continually, about once a month, because you will forget important details faster than you expect.
5. Ask your agent to review recent prompts, git history, Jira tickets, pull requests, docs, and chat threads, then add evidence-backed notes under the right role.
6. For each job, give verboseResume.md and the job description to your LLM.
7. Ask for upload JSON, preview on this site, then keep revising the JSON with the same bot until it reads right.
8. Print or save PDF when the preview looks good. Use browser edit mode only for tiny layout fixes.

Verbose resume Markdown format

Do not upload verboseResume.md to the site. It is your long memory; the upload file is concise JSON only.

Agents can fetch the full spec via MCP get_verbose_resume_format. In the repository, see docs/VERBOSE-RESUME.md and the example verboseResume.md.

# Your Name — Verbose Resume

## Contact
- **Email:** you@example.com

## Summary
Long-form narrative.

## Skills
### Languages
- Go

## Experience
### Company — Title
**Dates:** Apr 2024 – Present | **Location:** Remote

#### Overview
Role scope.

#### Notable work
##### Project name
Deliverables and tools.

Upload JSON format

The upload file is the concise final resume. It should match this shape:

{
  "Name": "Your Name",
  "Location": "City, ST",
  "Phone": "555-0100",
  "Email": "you@example.com",
  "Website": "https://example.com",
  "Github": "https://github.com/example",
  "Summary": "Short tailored summary.",
  "Skills": {
    "Languages": ["Go", "Python", "Bash"],
    "Cloud & Containers": ["Kubernetes", "AWS"],
    "CI / CD": ["GitHub Actions", "ArgoCD"]
  },
  "Experience": [
    {
      "Title": "Role title",
      "Company": "Company name",
      "Location": "Remote",
      "Date": "Apr 2024 - Present",
      "Details": ["Tailored bullet focused on the target role."]
    }
  ],
  "Education": [
    {
      "Degree": "Degree name",
      "School": "School name",
      "Location": "City, ST",
      "Date": "May 2009"
    }
  ],
  "Certifications": ["Certification name"]
}

Tailor with an LLM (JSON loop)

Keep one verboseResume.md as your long memory. For each application, work in a tight loop with ChatGPT, Claude, Gemini, or another assistant—the artifact you iterate on is upload JSON, not a second Markdown resume.

1. Attach or paste verboseResume.md, the job description, and the upload JSON shape from the section above (or MCP get_upload_json_format).
2. Ask the model for only valid upload JSON. Save it as a .json file.
3. Upload that JSON here, pick a template, and review the HTML preview.
4. Not happy with a bullet, emphasis, or length? Paste the current JSON back to the same chat with plain-language feedback. Ask for a revised JSON. Re-upload and preview again.
5. When it looks right, print or save PDF. Optional: WYSIWYG edit mode for small wording tweaks in the browser—then fix the JSON in your bot if you want the file to stay in sync.

Do not maintain a parallel Markdown copy of the tailored resume. The verbose Markdown file is the source of truth; the JSON is the tailored snapshot for this job.

First pass prompt:

You are helping me tailor my resume for the job description below.
Use only facts from my verbose resume (verboseResume.md).
Do not invent employers, dates, tools, credentials, outcomes, metrics,
citizenship status, clearance, or experience.

Select the most relevant experience, compress it, and output only valid JSON
matching the Verbose Resume upload format.

Prioritize relevance, truthful impact, short specific bullets, and a resume
that fits cleanly into one to two pages.

Inputs:
1. My verbose resume (attached).
2. The target job description (below).
3. The Verbose Resume upload JSON spec (below).
4. Any special constraints I add.

Output only the final JSON.

Revision pass prompt (after you preview on the site):

Here is my current tailored resume JSON and feedback on the preview.

Revise the JSON only. Keep all facts truthful and supported by my verbose resume.
Do not invent experience. Output only the complete updated JSON in the same
upload format—no commentary outside the JSON.

Feedback:
- [e.g. shorten summary, stronger platform bullets, drop older role]

Current JSON:
[paste JSON here]

MCP endpoint

There is a lightweight MCP-compatible JSON-RPC endpoint at /mcp. Guidance tools:

• get_resume_generator_guide
• get_verbose_resume_format
• get_upload_json_format
• get_llm_prompt_guide

Artifact tool for agents (optional, not the human loop above):

• create_resume_artifact turns upload JSON into standalone HTML, normalized JSON, and local PDF instructions for agent-driven workflows.

Humans normally iterate JSON in chat, upload here, and print from the browser. Agents can use MCP tools to read specs and emit JSON the same way.