Resume Tailor

Resume Tailor

Use this skill when a user wants to adapt the generic resume to a specific job posting.

For full recruiting sessions that include status refreshes, applications, recruiter/engineer outreach, prospecting, and dashboard refreshes, start with recruiting-pipeline and let it route into this skill for new role tailoring.

When the user wants to focus only on resume tailoring but still keep the surrounding recruiting flow organized, run:

python3 skills/recruiting-pipeline/scripts/build_daily_recruiting_plan.py --mode resume

Then tailor the requested role, update the tracker, and only continue into apply/outreach steps when the plan or user request calls for it.

Inputs

Gather these before editing:

1. A job description URL, or the pasted job description text
2. The target company name
3. The generic resume source in generic-resume/
4. The candidate metadata in generic-resume/README.md

If a link is provided, open it and extract:

• Title
• Company
• Seniority
• Must-have skills
• Nice-to-have skills
• Repeated keywords
• Domain signals such as startup, AI, frontend, backend, infra, product, leadership, or customer-facing work

Files and folders

• Generic source resumes live in generic-resume/
• Use generic-resume/resume.tex as the early-career source when a posting is explicitly new-grad, university-grad, associate, junior, SWE I, or otherwise early-career. This version includes Liam's graduation month/year (Dec 2024) and GPA, and labels both Caterpillar roles as internships.
• Use generic-resume/resume-general.tex when a posting is not explicitly early-career. This version omits graduation month/year and GPA, labels Caterpillar May 2023-Dec 2023 as Software Engineer, and labels only Caterpillar May 2022-August 2022 as Software Engineer Intern.
• Candidate metadata lives in generic-resume/README.md
• The canonical personal website is liamvan.dev; every tailored resume header should preserve \href{https://liamvan.dev}{liamvan.dev}
• The canonical Fantasy Wizard project URL is https://fantasysportwizard.com; when including the project, prefer linking the project title as \href{https://fantasysportwizard.com}{Fantasy Wizard} when space and formatting allow
• Set candidate_name: Your Name in that README so the scripts can name output folders and PDFs
• Treat generic-resume/README.md as the richer candidate profile and evidence source, not just naming metadata
• Treat the selected generic resume and README as the context bank; they can be richer than one page because the tailored output is what must be compressed to one page
• Company outputs should live in companies/<Company Name>/<Role_Slug>/Liam_Van_Resume/ when a role title is known for Liam's pipeline
• If the same role is tailored again, create companies/<Company Name>/<Role_Slug>/<Candidate_Name>_Resume_2, then _3, and so on
• Fall back to companies/<Company Name>/Liam_Van_Resume/ only when the role is unknown
• Application tracking lives in application-trackers/applications.md
• Optional Notion mirroring is handled by the separate notion-application-sync skill and should not run during normal resume tailoring.

Batch tailoring jobs live in skills/resume-tailor/config/tailor_jobs.json, and reusable profile text lives in skills/resume-tailor/config/skill_profiles.json. For sourced batches, append a job record to that manifest and run:

python3 scripts/tailor.py --batch "tailor_intake_YYYY_MM_DD"

Use python3 scripts/tailor.py --list-batches to see available batch ids and python3 scripts/tailor.py --validate-manifest to check generated resume folder/PDF paths recorded in the manifest.

Before editing, prepare the output directory with:

python3 skills/resume-tailor/scripts/prepare_resume_folder.py \
  --company "Company Name" \
  --role "Role Title"

This copies the generic resume files into the correct company and role specific folder and prints the destination path.

After editing, render a final PDF in the same folder with:

python3 skills/resume-tailor/scripts/render_resume_pdf.py \
  --dir "companies/Company Name/Role_Slug/Liam_Van_Resume"

This attempts to compile resume.tex and writes a final PDF named Liam_Van_<Company_Name>.pdf in that same company-specific folder.

After every resume PDF render, verify that the PDF is exactly one page and materially fills the page:

python3 skills/resume-tailor/scripts/verify_resume_pdf.py \
  --pdf "companies/Company Name/Role_Slug/Liam_Van_Resume/Liam_Van_Company_Name.pdf"

If verification fails because the resume spills past one page, trim or compress the least relevant content, rerender, and verify again. If verification fails because the one-page resume is visibly underfilled, add more truthful, role-relevant evidence from generic-resume/README.md, the generic resume, or the provided job context, then rerender and verify again. Do not update the tracker until the rendered PDF passes this check or you have manually inspected the PDF and can explain why the automated fill check is unavailable or too conservative.

If the user wants a basic cover letter in the same folder, create it with:

python3 skills/resume-tailor/scripts/create_cover_letter.py \
  --dir "companies/Company Name/Role_Slug/Liam_Van_Resume" \
  --company "Company Name" \
  --role "Role Title" \
  --why-interest "Two to three sentences on why the role is a fit"

Then render the cover letter PDF with:

python3 skills/resume-tailor/scripts/render_cover_letter_pdf.py \
  --dir "companies/Company Name/Role_Slug/Liam_Van_Resume"

This writes cover_letter.tex and a final PDF named Liam_Van_<Company_Name>_Cover_Letter.pdf right next to the tailored resume PDF. Do not leave or upload Candidate_Name_... artifacts.

After the folder and PDF exist, update the markdown tracker with:

python3 skills/resume-tailor/scripts/update_application_tracker.py \
  --company "Company Name" \
  --role "Role Title" \
  --job-link "https://example.com/job" \
  --location "City, State" \
  --source "Ashby" \
  --referral "No" \
  --date-added "YYYY-MM-DD" \
  --resume-folder "/absolute/path/to/companies/Company Name/Liam_Van_Resume" \
  --resume-pdf "/absolute/path/to/companies/Company Name/Liam_Van_Resume/Liam_Van_Company_Name.pdf" \
  --status "Resume Tailored"

The tracker also supports a configurable fit score and recruiter outreach flag:

• preferences live in application-trackers/scoring-profile.json
• Fit Score is a 1 to 10 heuristic based on role, location, company, source, and current status
• Reach Out should default to Yes when the score is at or above the configured threshold
• use --fit-score or --reach-out only when you need to override the defaults for a specific role
• when Reach Out is enabled, update_application_tracker.py also upserts the role into application-trackers/outreach-prospects.md so recruiter and engineer queue views include the new tailored application immediately

To backfill or recompute those values across the full tracker, run:

python3 skills/resume-tailor/scripts/score_application_tracker.py

Use the tracker every time the user provides a new role link so the repository keeps a single running application log. If the user mentions a referral, record it in the Referral column. If they do not mention one, leave it blank or mark No based on the user's preference. Treat the posting itself as the unique identity of the application. Multiple roles at the same company should create separate folder paths and separate tracker rows, even when the titles are similar. Use status values such as Resume Tailored, Applied, Online Assessment, Interviewing, Offer, Rejected, and Archived when they fit the user's application stage. Keep the markdown tracker header count in sync so application-trackers/applications.md shows Total applications tracked: N above the table. If the resulting row has Reach Out enabled and the user wants outbound networking, hand off to the linkedin-outreach skill next so recruiter and engineer outreach gets tracked in the same markdown row.

If the user wants Notion updated after tailoring, hand off to notion-application-sync after the markdown tracker and visualizer cache are refreshed.

Tailoring rules

Keep the resume truthful. Optimize framing, ordering, emphasis, and brevity. Do not invent tools, impact, titles, or employers.

Primary edits:

1. Reorder the skills section so the most relevant skills appear first
2. Remove weak or irrelevant skills that dilute the match
3. Add already-true skills that exist elsewhere in the resume or user-provided context if they improve alignment
4. Rewrite bullets to foreground matching technologies, outcomes, scope, and ownership
5. Drop or compress older or less relevant roles if needed to stay within one page
6. Preserve measurable impact whenever possible
7. Keep the final tailored result to exactly one page after LaTeX render
8. Use the full page effectively when strong truthful content fits; avoid leaving obvious empty space in the final one-page layout
9. Keep the bullet-based layout; do not replace experience bullets with paragraph blocks
10. Allow two-line bullets when they improve clarity, but do not force bullets to fill space unnecessarily
11. Keep the skills section visually substantial so it never looks sparse or underfilled
12. Preserve a compact header with the personal website liamvan.dev and no blank-looking contact rows
13. Default tailored role labels to Software Engineer unless the user explicitly wants the original internship wording
14. Avoid bullets that wrap with only one or two low-information trailing words on the second line; rewrite those bullets so wrapped lines carry meaningful content

Skill alignment workflow

Before editing the skills section, build a quick alignment map with four buckets:

1. Required or highly repeated skills from the job posting
2. Skills already listed in the resume
3. Missing-but-true skills implied by bullets, projects, or technologies named elsewhere in the resume
4. Low-value or irrelevant skills that should be removed or demoted

Use that map to drive edits instead of only reordering the existing list.

Rules for adding or removing skills

• Add a skill if the resume wording clearly supports it, even if the current skills section does not list it
• Infer missing-but-true skills from tools, frameworks, cloud platforms, testing systems, languages, and product areas mentioned in bullets or project lines
• If a bullet says the candidate built APIs with Spring Boot, it is valid to surface REST APIs and Spring Boot in the skills section
• If a bullet says the candidate used Robot Framework, OpenSearch, OCI, GCP, React, NestJS, Android, Selenium, or Azure DevOps, those can be promoted into skills when relevant
• Do not add a skill unless the resume already provides evidence for it
• Remove or de-emphasize generic skills that do not improve match quality for the target job
• Remove older, weaker, or redundant items when they crowd out better evidence
• Prefer fewer, stronger, more targeted skills over long catch-all lists
• When a role strongly implies a skill but the wording is weak, tighten the bullet language so the evidence is clearer before adding that skill

Editing workflow

1. Read the job description and extract the top priorities.
2. Inspect generic-resume/README.md, then select the right base: generic-resume/resume.tex for explicit early-career/new-grad/associate/junior/SWE I roles, or generic-resume/resume-general.tex for broader roles that do not need graduation date, GPA, or internship framing.
3. Read generic-resume/README.md in full and use it as the richer profile, evidence bank, and wording guide.
4. Use candidate_name from generic-resume/README.md for output naming.
5. Create the company-specific copy with prepare_resume_folder.py.
6. Edit the copied LaTeX files in the company-specific folder, not the generic source.
7. Build a skill alignment map:
   • top skills from the posting
   • skills already listed
   • missing-but-true skills found in bullets and projects
   • stronger evidence and wording available in generic-resume/README.md
   • skills to remove or demote
8. Make targeted changes:
   • skills ordering
   • adding missing-but-true skills
   • removing weak or irrelevant skills
   • pulling stronger evidence from generic-resume/README.md
   • rephrasing bullets with more specific truthful wording from generic-resume/README.md
   • summary wording if the template has one
   • bullet emphasis
   • removal of lower-value lines or roles
   • preserve readability over over-compression, even if that means a few bullets wrap to a second line
   • inspect wrapped bullets and rewrite any that leave an orphaned trailing word or tiny fragment on the next line
   • preserve the source-specific education/Caterpillar framing unless the job posting or user request clearly calls for switching variants
9. Check for one-page risk. If the resume is too long:
   • remove least relevant bullets first
   • shorten wording second
   • remove an older role only if necessary
   • before cutting information, try rewriting awkwardly wrapping bullets so the line break is cleaner
10. If the resume fits comfortably under one page with visible unused space:

• add stronger truthful detail from the generic context before considering layout-only expansion
• prefer strengthening bullets, skills, or a relevant project line over leaving the page underfilled
• keep the final rendered PDF to exactly one page

11. Render the final PDF with render_resume_pdf.py.
12. Run verify_resume_pdf.py on the rendered PDF.
    • If it reports more than one page, remove, compress, or tighten lower-priority content and rerender.
    • If it reports underfilled page usage, add truthful, role-aligned detail before rerendering.
    • If the automated fill check is unavailable, manually inspect the PDF before proceeding.
    • Repeat render and verification until the PDF is exactly one full page and no more than one page.
13. Update application-trackers/applications.md with the new role, posting key, link, resume folder, PDF path, source, referral status, and current status.
    • Include a company-and-role labeled PDF link so the tracker can open the tailored resume directly from the table.
    • Keep the markdown tracker ordered for fast scanning: Company, Role, Applied, Status, Company Resume, Referral, then supporting details.
    • Refresh the header count so the markdown tracker shows the current total number of tracked applications.
14. Refresh the visualizer cache when tracker rows changed.
15. If the user explicitly wants Notion updated, hand off to notion-application-sync.
16. Summarize what changed and why, including which skills were added, removed, promoted, or rewritten using stronger profile evidence, and state that the rendered PDF passed the one-page/full-page verification.

Prioritization heuristics

Bias toward:

• Keywords repeated in required qualifications
• Tools that appear in both requirements and responsibilities
• Skills tied to business outcomes, ownership, and cross-functional execution
• Recent experience over older experience
• Matching domain language such as AI, SaaS, data, platform, security, fintech, healthcare, or consumer
• Skills that are clearly evidenced in bullets even if they are missing from the current skills section
• Stronger supporting detail from generic-resume/README.md when the one-page resume is too compressed to show the best phrasing

De-prioritize:

• Generic skills that take space without improving relevance
• Older technologies unless the job explicitly asks for them
• Bullets that describe tasks without outcomes
• Experience that competes with the target narrative
• Skills with no clear evidence in the resume

Output expectations

The final deliverable should usually include:

1. Updated LaTeX resume files in the company-specific directory
2. A final PDF named Liam_Van_<Company_Name>.pdf in that same directory when local LaTeX tooling exists
3. When requested, a basic cover_letter.tex and Liam_Van_<Company_Name>_Cover_Letter.pdf in that same directory
4. An updated row in application-trackers/applications.md
5. A brief summary of changes made
6. Any notes about missing evidence for requested skills

If local LaTeX tooling is missing, leave the LaTeX ready for Overleaf upload and note the intended PDF filename.

Overleaf note

Direct Overleaf upload is not built into this skill. If the user has connected the project to Git, you can push the tailored LaTeX to that Git remote. Otherwise, prepare the company-specific LaTeX folder so it can be uploaded to Overleaf manually.