k-skincare

Evidence-based K-skincare + K-wellness consultation. RCT-grounded. Multi-language.

This repo ships two Claude Code skills + a standalone interactive CLI:

Skill	Scope
k-skincare	acne, pigmentation, dark circles, retinoid protocols, Korean derm procedures, cleanser pH, sensitive skin, body acne
k-wellness	supplements, sleep / phase advance, cutting / visceral fat, cortisol, hyperventilation, gut / flatus, eye fatigue

---

Quick start

Skills (Claude Code, Cursor, Cline, Copilot, 18+ AI agents)

# Pick both / either skill interactively
npx skills add seonglae/k-skincare

# List what's in the repo
npx skills add seonglae/k-skincare --list

# Install one skill non-interactively to global Claude Code
npx skills add seonglae/k-skincare --skill k-skincare  -g -a claude-code -y
npx skills add seonglae/k-skincare --skill k-wellness  -g -a claude-code -y

After install, restart Claude Code (or /skills refresh). The skill auto-activates when the user asks about skincare or wellness.

Powered by Vercel Skills. Listed at skills.sh/seonglae/k-skincare once install telemetry reaches the threshold.

Interactive CLI (no Claude required)

npx k-skincare

Pastel + Ink terminal wizard. Asks age / sex / region (UK/KR/US/EU) / skin type / concerns / budget → returns AM/PM routine + region-specific shopping list + RCT citations + warnings (drug interactions, cleanser pH, laser wavelength choice). English. Rule-based engine. No API cost.

---

k-skincare skill

Triggers when user mentions any of:

• Skin concerns — acne, dark circles, pigmentation, pores, wrinkles, sensitivity, scars
• Products — cleanser, toner, serum, moisturizer, SPF, retinoid, AHA/BHA
• Routine building — AM, PM, weekly, retinoid ramp-up
• Korean derm procedures — pico laser, Nd:YAG, PDL, IPL, fractional, MTS

Behavior:

1. Detects user language from first message — conducts entire conversation in that language
2. Tier 1 intake (5 essential questions) → basic plan
3. Offers Tier 2 (10 refinement questions) → precise plan
4. Tier 3 (procedure planning) if user is open to in-clinic
5. Output — personalized AM/PM routine + retinoid ramp-up + region-specific shopping list + stop conditions + RCT citations

Evidence base — skills/k-skincare/evidence/ (6 peer-reviewed summaries):

• Retinoid efficacy (Shalita 1996, Bagatin 2018, Kang 2005)
• Cleanser pH (Korting 1995, Gfatter 1997)
• Truncal Malassezia folliculitis (Paichitrojjana 2022)
• Solar lentigines / pico laser (Vachiramon 2022, Negishi 2018)
• Azelaic acid for PIE/PIH (Liu 2024, Thiboutot 2003)
• Dark circles laser selection (AlRamthan 2024 — 1064nm Nd:YAG vs 532nm KTP)

Refers out:

• Severe nodulocystic acne → derm for isotretinoin
• Suspicious lesions → urgent derm
• Pregnancy / breastfeeding + retinoid → derm / OB-GYN
• Mental health (BDD, picking) → GP / therapy

---

k-wellness skill

Triggers when user mentions any of:

• Supplements — NAC, magnesium, omega-3, vitamin D, K2, ashwagandha, L-theanine, PS, lutein, astaxanthin, glycine, biotin, zinc
• Sleep — delayed sleep phase, sleep onset insomnia, melatonin, Circadin, Slenyto, phase advance, morning grogginess, REM, deep sleep
• Diet / weight — cutting, visceral fat, BMR/TDEE, caloric deficit, time-restricted eating, protein target, sulfur load, FODMAP
• Mental / stress — hyperventilation, panic, anxiety baseline, cortisol regulation, HPA axis, CBT, Buteyko
• GI — flatus odor, bloating, malabsorption, probiotics, FOS, RS (resistant starch), enzymes
• Vision (screen) — blue light, lutein, astaxanthin, 20-20-20 protocol

Behavior — same Tier 1/2/3 intake pattern. Always checks current Rx list before any supplement recommendation. Always offers behavioral first-line when evidence supports.

Evidence base — skills/k-wellness/evidence/ (6 peer-reviewed summaries):

• Diet → stress (Lassale 2019 SR/MA, SMILES Jacka 2017, Su 2018 JAMA omega-3 anxiety)
• Diet → sleep (Sletten 2018 melatonin DSWPD, Drake 2013 caffeine, Yamadera 2007 glycine PSG)
• Diet → eye fatigue (Stringham 2017 macular carotenoids, Nagaki 2002 astaxanthin, DREAM 2018 NEJM negative for omega-3 capsules)
• Diet → gut / gas (Magee 2000 protein → H₂S, Yao 2018 RS suppresses H₂S 82-89%, Suarez 1998 bismuth)
• Ashwagandha (Chandrasekhar 2012 cortisol -27.9%, NIH LiverTox C 2024 hepatotoxicity)
• Vitamin D3 + K2 (Hyppönen 2007 UK 87% deficient, Park 2018 KNHANES Korean 20-34M cohort)

Refers out:

• Diagnosed mental illness needing meds → GP / psychiatrist
• Suspected eating disorder → NEDA / BEAT + GP
• Suspected sleep apnea → GP for polysomnography
• Severe vitamin deficiency requiring blood tests → GP
• Pregnancy / breastfeeding supplement decisions → OB-GYN
• Pediatric (<18) → pediatrician
• Hormonal therapy (testosterone, thyroid) → endocrinologist

---

Language support

Korean / English / Japanese / Chinese / Spanish (tested patterns). Other languages: AI runtime translation. Technical terms (Adapalene, NAC, KSM-66, mg, EPA, DHA) kept in original with brief native gloss.

What the skill produces

Both skills follow the same output structure:

1. Summary — patient profile + top 3 concerns + overall strategy
2. Concern → Intervention mapping — table with OTC / Rx / in-clinic columns, RCT citations
3. AM routine table — step-by-step with product type + dose + technique
4. PM routine table — same; for skincare, alternating days for retinoid + AHA scheduling
5. Active ramp-up schedule — W1-2 / W3-4 / W5+ frequency
6. Shopping list — priority + region cost estimate
7. Stop conditions — safety triggers per intervention
8. Evaluation checkpoints — W4 / W8 / W12 photo + self-score
9. Refer-out flags — if applicable
10. Evidence citations — first-author, year, journal, effect size

Anti-marketing principles

• Only RCT/SR-evidenced interventions get recommended; expert-consensus items are labeled as such with reasoning
• Effect sizes only — never "boosts", "rejuvenates", "detoxifies"
• Specific dose + timing + duration — never "use as needed"
• Acknowledges where evidence is weak (Vit K2 in young adults, blue-light glasses for sleep — both rejected per evidence)

Project structure

.
├── skills/
│   ├── k-skincare/
│   │   ├── SKILL.md
│   │   ├── intake.md             (Tier 1/2/3 multi-language form)
│   │   ├── references/
│   │   │   ├── decision-trees.md (concern → intervention with RCT anchors)
│   │   │   ├── output-format.md  (10-section plan standard)
│   │   │   └── language-handling.md
│   │   └── evidence/             (6 RCT summaries)
│   └── k-wellness/
│       ├── SKILL.md
│       ├── intake.md
│       ├── references/           (same 3 files, adapted)
│       └── evidence/             (6 RCT summaries)
├── cli/                          (Pastel + Ink + TS — interactive terminal CLI)
├── bin/                          (n/a — single bin via package.json)
├── README.md
├── LICENSE
└── package.json                  (single npm package: `k-skincare`)

Contributing

Issues and PRs welcome. Open an issue first for substantive changes. RCT additions/corrections especially appreciated — must include PMID/DOI.

License

MIT. Cite RCT primary sources, not this skill.

---

Disclaimer: This software provides general skincare / wellness information based on peer-reviewed research. It is NOT medical advice. Consult a qualified dermatologist / physician for individual diagnosis and treatment.