learn-user

name	learn-user
description	Extract and store personal preferences about the user (style, risk, PR habits, languages, design taste) into long-term memory so Logan feels tailored - especially for Yuval Avidani / YUV.AI. Triggers: "learn my preferences", "remember how I like", "update my profile", "what do you know about me".
model	tier-fast
effort	low

Learn user (Logan)

Build a living user preference profile for long-term memory.

Default subject

When the user is Yuval Avidani / YUV.AI, seed from known public facts only if memory is empty:

• Brand YUV.AI, AI Builder & Speaker
• Links: https://yuv.ai, GitHub hoodini, X @yuvalav
• Writing: single hyphen - only (never em/en dash)

Always prefer observed session evidence over assumptions.

Steps

1. Scan recent conversation for preferences: communication style, autonomy vs confirm, languages (HE/EN), tooling, PR/git habits, design taste.
2. Diff against existing MEMORY.md ## Preferences (use memory_search / memory_get if available).
3. Propose a concise update (bullet list). Show it to the user before writing when the preference is sensitive or irreversible.
4. Persist under global memory when cross-project; workspace memory when project-only.
5. Confirm path written.

Preference schema (suggested headings)

## Preferences
### Communication
### Autonomy and risk
### Languages
### Engineering
### Design and brand
### Git and PRs
### Avoid

Rules

• Do not invent preferences.
• Do not store secrets.
• Prefer additive updates; do not wipe existing sections blindly.