@getkrafter/resume-toolkit

Deterministic resume scoring, ATS keyword matching, and guided resume improvement. Works as an MCP server, a scoring library, and an AI skill for Claude Code, Cursor, and Windsurf.

Same input always produces the same score. The AI handles parsing and explanation — the scoring is pure math.

Quick Start — Skill (no account needed)

All platforms (Claude Code, Cursor, Windsurf, OpenCode, Codex)

npx skills add getkrafter/resume-toolkit

Then use /score in any conversation to start.

What You Can Do

Score your resume for general quality

Just run /score and paste your resume when asked. You'll get a breakdown across 5 dimensions — measurable results, action verbs, bullet quality, section coverage — with specific before/after suggestions for your weakest areas.

/score
> [paste your resume]
> [skip the JD prompt]

Score against a specific job description

Provide a JD to unlock ATS keyword matching. You'll see which keywords you're hitting, which are missing, and exactly which bullets to tweak to close the gaps — with confidence notes so you never add something you didn't actually do.

/score
> [paste your resume]
> [paste the job description]

Score from a file

Point to a file on disk instead of pasting:

/score
> my resume is at ~/Documents/resume.pdf

Score a Krafter resume

If you have the MCP server configured with a Krafter API key, pull your resume directly:

/score
> score my Krafter resume

Walk through every bullet with a guided interview

After seeing your scores, say "yes" when asked if you'd like to improve every bullet. The skill walks you through each one, asking for your real numbers — team sizes, user counts, performance gains — one question at a time. You pick from suggested ranges or type your own answer. It then crafts the improved bullet using your actual experience.

This is the core value: you don't need to be a resume writer. Just answer the questions honestly and the skill frames your experience in the strongest truthful way.

Quick Start — MCP Server

Without API key (scoring tools only)

{
  "mcpServers": {
    "krafter": {
      "command": "npx",
      "args": ["@getkrafter/resume-toolkit"]
    }
  }
}

With API key (scoring + Krafter CRUD)

{
  "mcpServers": {
    "krafter": {
      "command": "npx",
      "args": ["@getkrafter/resume-toolkit"],
      "env": {
        "KRAFTER_API_KEY": "sk-your-key-here"
      }
    }
  }
}

Generate an API key at krafter.vercel.app → Settings → AI Integrations.

Library Usage

npm install @getkrafter/resume-toolkit

import { scoreResume, scoreATS, toResumeData } from '@getkrafter/resume-toolkit';

// Score a resume
const result = scoreResume(
  { rawText: '...', bullets: ['...'], sections: ['experience', 'skills'] },
  'Job description text...' // optional
);

console.log(result.total);     // 0-100
console.log(result.mode);      // 'with-jd' or 'without-jd'
console.log(result.breakdown); // per-dimension scores
console.log(result.flags);     // diagnostic messages

// ATS keyword match only
const ats = scoreATS(resumeText, jdText);
console.log(ats?.matched);  // keywords found
console.log(ats?.missing);  // keywords missing

// Convert Krafter resume object to scoreable format
const resumeData = toResumeData(krafterResumeObject);

MCP Tools

Public (no auth)

| Tool | Description | |---|---| | score_resume | Full quality score (0-100) across 5 dimensions with breakdown and flags | | score_ats | ATS keyword match with bigram/unigram analysis |

Krafter (API key required)

| Tool | Description | |---|---| | score_krafter_resume | Fetch resume from Krafter → score in one call | | get_resume | Fetch a resume by ID | | list_resumes | List all your resumes | | create_resume | Create a new resume | | update_resume | Update an existing resume | | delete_resume | Delete a resume | | duplicate_resume | Clone a resume | | update_settings | Update visual settings | | update_section | Update a specific section | | list_templates | List available templates | | get_resume_schema | Get the resume data schema |

Scoring Dimensions

| Dimension | Weight (with JD) | What it measures | |---|---|---| | Quantification | 25% | Fraction of bullets with numbers/metrics | | Verb Strength | 20% | Quality of action verbs (tier1 > tier2 > tier3) | | ATS Match | 30% | Keyword overlap with job description | | Bullet Structure | 15% | Verb + number + detail pattern | | Section Completeness | 10% | Presence of expected resume sections |

When no JD is provided, the 30% ATS weight redistributes proportionally across the other dimensions.

License

MIT