LaTeX CV Refresh Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Produce an image-free, date-specific A4 CV with current 2026–2027 publication metadata, four books, verified links, and a reproducible Overleaf package.

Architecture: Keep the existing LaTeX section structure and wrapping-safe commands. Add a small source-contract test suite for content corrections, then compile and visually inspect the actual PDF; the live P1 WEB_Publications snapshot supplies baseline metadata and explicit user corrections take precedence.

Tech Stack: LaTeX/pdfLaTeX, latexmk, Python unittest, Poppler rendering, ZIP.

Spec: docs/superpowers/specs/2026-09-28-lectures-cv-publications-design.md

Global Constraints

  • Work in /data/khkim/1_users/3_stair123/1_projects/01_git_homepage_projects/04_latex_mycv/02_temp/swe_khkim_template_260928.
  • Final PDF must be A4 and contain no overfull boxes.
  • Teaching and volunteering sections contain no images.
  • Do not invent paper URLs; leave titles unlinked when no verified PDF exists.
  • User corrections override the live sheet.
  • Preserve author ordering and bold Kyeonghun Kim.

Review Focus

  • URL characters such as #, _, and & compile without breaking hyperref.
  • Long titles and author lists wrap inside A4 width.
  • A moved paper appears once, under the correct year and venue.
  • Removed HERO title is absent from both source and extracted PDF text.
  • Final ZIP compiles in a clean directory without relying on files outside the package.

Task 1: Expand teaching and remove CV images

Files:

  • Create: tests/test_cv_content.py
  • Modify: src/lectures.tex
  • Modify: src/volunteering.tex
  • Modify: resume_khkim.tex only if image-only packages become unused.

Interfaces:

  • Produces: seven dated teaching entries and two text-only volunteering entries.

  • Step 1: Write failing source-contract tests

Assert these seven exact date/organization pairs, absence of includegraphics in both files, absence of the grouped Professional AI Training heading, and retention of Red Cross/NAVER links:

  • foundational university-level instruction, 2018–2026;
  • SK Telecom, Dec. 5–7, 2023;
  • SNU OUTTA AI Bootcamp, Jun. 2024–Aug. 2026;
  • HUN Company, Apr. 2025;
  • KCA / Kyobo Life, Apr. 2026;
  • Busan Metropolitan City, Jul. 2026;

  • Kookmin University AX-Startup Bootcamp, Aug. 29 and Sep. 1, 2026.

  • Step 2: Verify RED

Run: python3 -m unittest tests.test_cv_content.CVContentTests.test_teaching_is_dated_and_all_resume_images_are_removed -v

  • Step 3: Rewrite the two sections

Use one compact resumeProjectHeading per teaching engagement. Remove every image block while preserving concise achievement text.

  • Step 4: Run focused test and compile

Run: python3 -m unittest tests.test_cv_content -v && latexmk -pdf -interaction=nonstopmode -halt-on-error resume_khkim.tex

Expected: tests pass and PDF compiles.

Task 2: Rebuild 2026–2027 publications

Files:

  • Modify: tests/test_cv_content.py
  • Modify: src/publications.tex

Interfaces:

  • Consumes: live P1 publication snapshot and the explicit corrections in the spec.

  • Produces: a deduplicated 2026–2027 LaTeX publication list with paper, venue, code, poster, and slide actions.

  • Step 1: Add failing metadata tests

Assert all corrected title/venue/status pairs, both supplied NeurIPS Drive IDs, AACL acceptance, HERO absence, and the three exact AICAS slide IDs. Assert Junsu Lim and Yului Jeong author URLs.

  • Step 2: Verify RED

Run: python3 -m unittest tests.test_cv_content.CVContentTests.test_2026_2027_publications_match_verified_metadata -v

  • Step 3: Replace the 2026–2027 blocks

Use current sheet metadata for other papers. Apply the explicit overrides and author links. Link venues to official conference pages and titles only to verified PDFs/arXiv pages.

  • Step 4: Run tests, compile, and inspect PDF text

Run: python3 -m unittest tests.test_cv_content -v && latexmk -pdf -interaction=nonstopmode -halt-on-error resume_khkim.tex && pdftotext resume_khkim.pdf - | rg 'ICML 2027|NeurIPS 2026|AACL|ISBI 2027|ICLR 2027'

Expected: all corrected venues appear; HERO does not.

Task 3: Add four books and linked poster coauthors

Files:

  • Modify: tests/test_cv_content.py
  • Modify: src/other_publications.tex

Interfaces:

  • Produces: Books and Conference Posters subsections inside Other Publications.

  • Step 1: Add failing book and author-link tests

Assert The Great White Drive ID, pen name Sunyul Baek, FBI dataset note, three exact English textbook titles, Jan. 2027, Hongneung Science Publishing, OUTTA use, and Junsu/Yului URLs.

  • Step 2: Verify RED

Run: python3 -m unittest tests.test_cv_content.CVContentTests.test_books_and_other_publications_are_complete_and_linked -v

  • Step 3: Rewrite Other Publications

Create compact year/subsection labels. Link every author for whom a verified personal URL is available without linking unknown authors.

  • Step 4: Run all CV tests and compile twice

Run: python3 -m unittest discover -s tests -v && latexmk -pdf -interaction=nonstopmode -halt-on-error resume_khkim.tex

Expected: tests and compilation pass.

Task 4: Visual QA and package delivery

Files:

  • Replace: Kyeonghun_Kim_CV.pdf
  • Replace: Kyeonghun_Kim_CV_Overleaf.zip
  • Modify: README.md only if build instructions change.

Interfaces:

  • Produces: self-contained A4 PDF and Overleaf ZIP.

  • Step 1: Verify layout diagnostics

Run pdfinfo, scan resume_khkim.log for overfull boxes, and render every page with pdftoppm.

  • Step 2: Visually inspect every rendered page

Check clipping, isolated headings, crowded entries, image absence, and balanced whitespace. Adjust only layout settings needed to correct observed defects, recompiling after each change.

  • Step 3: Rebuild artifacts

Copy the final PDF to Kyeonghun_Kim_CV.pdf. Create Kyeonghun_Kim_CV_Overleaf.zip from README.md, LaTeX sources, required assets, and the final PDF, excluding auxiliary files and removed teaching/volunteering images.

  • Step 4: Clean-room validation

Extract the ZIP into a fresh mktemp -d directory, compile there, verify A4 dimensions and no overfull boxes, then remove only that temporary directory.

  • Step 5: Final P1 verification and push

In P1 run python3 -m unittest discover -s tests -v, git diff --check, fetch origin, verify no remote divergence, and push main. Confirm the official Pages workflow and live /CV, /lectures, and /publication routes.