Resume Tailor

• Extract text from the current resume (use pandoc for .docx, pdf-reading skill for .pdf)
• Extract text from the job description

2. Analyze the job description

Extract and categorize:

• Hard skills — languages, frameworks, tools, certifications (e.g. "Python", "AWS", "PMP")
• Soft skills — leadership, communication, collaboration patterns
• Domain keywords — industry terms, methodologies (e.g. "Agile", "CI/CD", "SaaS")
• Seniority signals — years of experience, scope of impact, team size
• ATS-critical phrases — exact phrases from the JD that applicant tracking systems will scan for

3. Audit the current resume

• Map which JD keywords already appear in the resume
• Identify gaps — JD keywords with no resume match
• Identify buried strengths — relevant experience that exists but is understated
• Flag irrelevant filler — bullets that add no value for this specific role

4. Rewrite the resume

Apply these rules in order:

a) Professional summary (top of resume)

• Write a 2–3 sentence summary that mirrors the JD's core requirements
• Front-load the most relevant title, years of experience, and domain
• Weave in 3–5 high-priority keywords naturally

b) Experience bullets

• Rewrite each bullet using the XYZ formula: "Accomplished [X] as measured by [Y], by doing [Z]"
• Inject missing JD keywords where the candidate has genuine experience — never fabricate
• Lead with strong action verbs matching the JD's tone (e.g. "Architected" for senior roles, "Implemented" for mid-level)
• Quantify impact wherever possible (%, $, time saved, users affected)
• Reorder bullets within each role: most JD-relevant first

c) Skills section

• Reorganize to lead with skills the JD prioritizes
• Add skills the candidate demonstrably has but omitted
• Group logically (Languages, Frameworks, Cloud/Infra, Tools, Methodologies)

d) Section ordering

• Place the most JD-relevant sections highest
• For senior roles: Summary → Experience → Skills → Education
• For technical roles: Summary → Skills → Experience → Education
• Remove or minimize sections that don't serve this application

e) General polish

• Eliminate first-person pronouns
• Keep to 1 page if <8 years experience, 2 pages max otherwise
• Use consistent date formatting (Mon YYYY)
• Ensure no orphan bullets (every role has 2–5 bullets)

5. Generate the output

Ask the user which format they want if not already specified. Then follow the appropriate path:

Option A: DOCX (default — best for ATS)

Use the docx skill (docx-js via Node.js) to produce a clean, ATS-friendly document:

• Font: Calibri or Arial, 10–11pt body, 12–14pt name
• Margins: 0.5–0.75 inches (tight but readable)
• No tables for layout — ATS parsers choke on tables; use paragraphs, tab stops, and spacing
• No headers/footers for contact info — some ATS systems skip header content
• No images, icons, or graphics — ATS can't read them
• Section headings: Bold, slightly larger, with a bottom border (Paragraph border, not a table)
• Bullet style: Use LevelFormat.BULLET from docx-js numbering config
• File name: {FirstName}_{LastName}_Resume.docx

Option B: PDF

Generate the resume as a .docx first using Option A rules, then convert:

python scripts/office/soffice.py --headless --convert-to pdf {FirstName}_{LastName}_Resume.docx

• File name: {FirstName}_{LastName}_Resume.pdf

Option C: Markdown

Output a clean .md file with:

• # for the candidate's name, ## for section headings
• Standard markdown bullet lists for experience items
• **Bold** for company names and job titles
• A horizontal rule (---) between major sections
• Contact info on a single line at the top using pipes: email | phone | location | LinkedIn
• File name: {FirstName}_{LastName}_Resume.md

6. Present results

Output to the user:

1. The tailored resume file in their chosen format
2. A brief changelog summarizing key changes:
   • Keywords added
   • Bullets rewritten (before → after for top 3)
   • Sections reordered
   • Gaps that couldn't be addressed (honest about what's missing)

Rules

• Never fabricate experience. Only surface and reframe what the candidate actually has.
• Never copy JD language verbatim into bullets. Paraphrase naturally.
• Preserve the candidate's voice. Match their existing tone and complexity level.
• ATS first, human second. The resume must pass automated screening before a human reads it.
• If the resume is a poor match (<30% keyword overlap), say so. Suggest which gaps to close before applying.