kjgarza generate-kristian-profile
.claude/skills/generate-application/references/kristian-profile.md is the candidate profile that the generate-application skill reads to score jobs and write CVs. It is a generated artifact — the source of truth is the site data under src/_data/ and the case studies under src/work/. This skill regenerates it so the two never drift.
All paths below are relative to the project root (kjgarza.github.io/).
Run (agent path)
The driver is .claude/skills/generate-kristian-profile/driver.mjs. It reads the data modules, parses the case studies, and writes the profile.
Preview without touching the file (prints full markdown to stdout — diff it against the current profile first):
node .claude/skills/generate-kristian-profile/driver.mjs --check | head -40
Write it for real:
node .claude/skills/generate-kristian-profile/driver.mjs
Expected output:
profile written → .claude/skills/generate-application/references/kristian-profile.md (13.9 KB)
sources: 7 case studies, 19 publications, 7 side projects, 42 skills
Write somewhere else (e.g. to compare):
node .claude/skills/generate-kristian-profile/driver.mjs -o /tmp/profile-new.md
What it pulls from where
| Profile section | Source |
|---|---|
| Identity (name, role, ORCID, languages) | src/_data/site.js (author, schema_org) |
| Positioning statement | src/_data/site.js (description, bio) |
| Employment history | src/_data/cv.js (employment) |
| Case study proof points (summary, highlights, metrics, tech) | src/_data/work.js order → each src/work/<slug>.md |
| Technical stack (bucketed) | src/_data/cv.js (skills) |
| Publications (grouped by theme) | src/_data/publications.js |
| Side projects | src/_data/tools.js |
| Education | src/_data/cv.js (education) |
Case study extraction keys off the case studies' consistent headings: ## Overview → summary, ### Technology Stack → tech, ## Key Achievements → highlights, bold numeric spans (**70%**, **50,000-dataset**) → metrics.
After running
Regenerating overwrites any hand edits — the file carries a "do not hand-edit" banner. If the profile needs a fact that isn't in the data, add it to the source (src/_data/*) or the driver, not the output file.
Two sections are not data-derived — they're positioning decisions, kept as constants at the top of driver.mjs: TARGET_ROLE (seniority, arrangement, Track A/B) and POSITIONING_FRAMES (frame → lead-evidence table). Edit those constants to change them; they survive regeneration.
Gotchas
- Case study body is prose, not structured fields. Highlights come from the
## Key Achievementsbullets; metrics come only from bold numeric spans. A case study with no## Key Achievementssection (e.g.redesigning-datacite-harvesting-services.md,creating-a-design-system.md) yields a Summary-only block — that's expected, not a failure. To enrich it, add a## Key Achievementslist to the.md. - The bold-span metric regex is single-line by design.
**...**spanning a newline is ignored — an earlier version captured multi-paragraph blobs. If a real metric goes missing, check it's bold and on one line in the source. --checkprints, does not write. Always the safe first move; the file is only touched by the argument-less run.gray-mattermust resolve. It's an Eleventy dependency, present afterbun install. If missing:bun install.
Troubleshooting
| Symptom | Fix |
|---|---|
Cannot find module 'gray-matter' |
bun install (it ships with Eleventy) |
Cannot find module '.../src/_data/cv.js' |
Run from the project root, or check the file exists |
| A case study block is missing highlights | The .md has no ## Key Achievements section — add one |
Skill name shows under Other in Technical Stack |
Add a keyword match to the buckets map in driver.mjs |