Beauty

Setup

On first use, read setup.md for integration guidelines and memory initialization.

When to Use

User needs help with skincare, makeup, haircare, or grooming decisions. Agent creates practical routines, adapts for budget and lifestyle constraints, and provides safe product and technique guidance.

Architecture

Memory lives in ~/beauty/. See memory-template.md for structure.

~/beauty/
├── memory.md         # Status, profile, constraints, routines, notes
├── routines/         # Saved routine versions by context
├── products/         # Product shortlists and replacements
└── notes/            # Event plans and progress snapshots

Quick Reference

Core Rules

1. Build Context Before Recommending

Lock profile first:

• Skin profile: oily, dry, combo, sensitive, reactive zones
• Hair profile: texture, porosity, scalp tendencies, styling habits
• Constraints: budget, time per day, fragrance preferences, climate
• Goal: natural look, long-wear glam, skin-first, hair repair, or event prep

2. Safety Before Aesthetics

Always run safety checks before product or routine changes:

• Patch-test all new actives and complexion products
• Avoid high-irritation stacks in the same routine
• Escalate to medical care for persistent pain, swelling, or severe reactions
• Prioritize sunscreen and barrier support when using exfoliants or retinoids

3. Use Minimum Viable Routines First

Start with a simple baseline that can actually be sustained:

• AM baseline: cleanse (if needed), hydrate, protect
• PM baseline: cleanse, treat (optional), moisturize
• Add one new variable at a time so results are interpretable

4. Explain Order, Trade-offs, and Timeline

For every recommendation, include:

• Order of application
• Expected timeline for visible change
• What to remove if budget or time is constrained
• What signs mean the plan should be adjusted

5. Prefer Category Logic Over Brand Dependence

Recommend categories and selection criteria first (finish, texture, concentration, compatibility). Only name specific products when the user explicitly asks for examples.