القائمة
"Analyze uploaded images using Gemini Vision, or generate/edit images with Gemini AI. Use for: 이미지 분석, 사진 분석, 이미지 생성, 그려줘, 그림 만들어줘, image generation, photo analysis"
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
One-shot import of Google Contacts (People API) into the user's Connect (인맥) list. Pages through every contact, dedupes against existing connections by email then phone, and inserts new rows with category=acquaintance and tags=['google_contacts']. Honors BR-SOCIAL-3 — never writes social_profiles even if Google returns LinkedIn URLs. Use for: 구글 연락처 가져오기, 연락처 동기화, contacts import, import google contacts, sync google contacts, google 연락처, 인맥 가져오기, 주소록 가져오기
Record and sync Connect (인맥) activity timeline entries. Pulls from Google Calendar + Gmail (shares google-workspace OAuth token), or accepts manual entries like '김철수와 어제 점심 먹었어'. Use for: 인맥 활동, 활동 기록, 만남 기록, 미팅 기록, 점심, 저녁, 통화 기록, 일정 동기화, 인맥 동기화, connect sync, connect activity, meeting log, met with, had lunch with
Add or replace Context Memo (context_notes) for a Connect (인맥) entry. Looks up the target connection by exact or fuzzy name when the agent doesn't already know the connection UUID. Use for: 인맥 메모, 인맥 메모 추가, 메모 추가, 메모 수정, 인물 메모, 사람 메모, connect memo, add note, update note, context note
Scan a business card image with Gemini Vision and extract structured contact fields (name, role, company, email, phone, address, etc.) to create a new Connect (인맥) entry. Use for: 명함, 명함 스캔, 명함 등록, 명함 인식, 명함 저장, 명함 추가, 인맥, 인맥 추가, 인맥 등록, 연락처 등록, 연락처 추가, 비즈니스 카드, business card, business card ocr, scan business card, contact card, register contact
Search the internet for up-to-date information and fetch web page content via Tavily. Use for: 검색해줘, 찾아줘, 최신 정보, 인터넷 검색, 뉴스, 요즘, web search, latest news, current information
Record notes in the Planner's '노트' section. Use for: 노트, 노트에 기록, 메모, 메모해줘, 노트 저장, 노트 추가, note, memo
| name | image |
| description | "Analyze uploaded images using Gemini Vision, or generate/edit images with Gemini AI. Use for: 이미지 분석, 사진 분석, 이미지 생성, 그려줘, 그림 만들어줘, image generation, photo analysis" |
Uses python3 image/scripts/analyze.py for image analysis and python3 image/scripts/image.py for image generation.
Always pass --user-id {user_id}.
GEMINI_API_KEY environment variable when configured. Always attempt to run the script — it will report if credentials are missing.DATABASE_URL, GATEWAY_URLMINIO_ENDPOINT, MINIO_ACCESS_KEY, MINIO_SECRET_KEYWhen to use: The message contains [image:URL] — meaning the user uploaded one or more images.
For each [image:URL] found in the message, call the analyze script once per image.
python3 image/scripts/analyze.py \
--user-id {user_id} analyze \
--file-url "{image_url}" \
--query "{user's question about the image}"
The script prints the Gemini Vision analysis result to stdout. Use that result to answer the user.
After receiving the analysis text, determine the image type and respond accordingly:
| Type | Indicators | Action |
|---|---|---|
| Receipt (영수증) | 금액, 날짜, 상호명 | Extract expense → offer to save to 가계부 |
| General photo | 풍경, 사람, 물체 | Describe the content |
| Text/Document | 글자, 표, 문서 | OCR and summarize |
| Screenshot | UI, 앱 화면 | Describe the screen content |
If the image is a receipt:
save_finance immediatelyIf the message contains multiple [image:URL] markers, analyze each one separately and combine the results.
User: "이 사진 뭔지 분석해줘 [image:http://localhost:9000/starnion-files/users/abc/2026/img.webp]"
python3 image/scripts/analyze.py \
--user-id abc \
analyze \
--file-url "http://localhost:9000/starnion-files/users/abc/2026/img.webp" \
--query "이 이미지를 자세히 분석해주세요."
User: "이 영수증 기록해줘 [image:/api/files/users/abc/2026/receipt.webp]"
python3 image/scripts/analyze.py \
--user-id abc \
analyze \
--file-url "/api/files/users/abc/2026/receipt.webp" \
--query "영수증의 가게명, 날짜, 총 금액, 항목을 추출해주세요."
When to use: User explicitly requests image creation ("그려줘", "만들어줘", "생성해줘").
python3 image/scripts/image.py \
--user-id {user_id} generate \
--prompt "{description in English}" \
--aspect-ratio 1:1
Aspect ratios: 1:1 (square), 16:9 (landscape), 9:16 (portrait), 4:3, 3:4
The script prints the image URL and a Markdown embed. Always include it in the response:
User: "고양이가 우주복을 입은 그림 그려줘"
python3 image/scripts/image.py \
--user-id abc123 generate \
--prompt "A cute cat wearing a spacesuit floating in space, digital art style" \
--aspect-ratio 1:1
--promptgemini-3.1-flash-image-preview (same as generation)gemini-3.1-flash-image-preview