Cloud Storage와 Claude Code로 관광 사이트 이미지 정리하기

관광 사이트는 계절 사진, 객실, 음식, 행사, 직원 촬영, 외부 작가 이미지를 한 폴더에 두기 쉽습니다. 다음 담당자는 어떤 사진이 공개 가능하고, 만료일이 있고, 크레딧이 필요한지 알기 어렵습니다.

이 글은 Cloud Storage를 검토 가능한 이미지 workflow로 바꿉니다. Claude Code는 명명 규칙, metadata 필드, alt 문구, 체크리스트를 만듭니다. 사람은 권리, 인물 노출, 크레딧, 공개 범위, 브랜드 판단을 봅니다.

이 글의 요점

• Start from a tourism page, image folder, and rights memo, not from a generic storage tutorial.
• Cloud Storage object names, metadata, public scope, alt text, and reviewer need one registry.
• Claude Code drafts names, metadata, alt text, and checklists; humans confirm legal and brand decisions.
• The main CTA is training because image rights and publishing workflow are site-specific.
• Measure ROI with image replacement time, missing alt text, rights review misses, and booking clicks.

실패 장면

공식 문서는 object metadata, public access prevention, static website hosting, Google Images SEO, image license metadata를 확인하는 기준입니다. Use Cloud Storage object metadata, Cloud Storage public access prevention, Cloud Storage static website hosting, Google Images SEO best practices, Google image license metadata as anchors. The platform can store and serve files, but it cannot decide if a tourism photo is licensed for the website, advertising, or social media.

The common failure is treating a file in a bucket as a file approved for publication. A better workflow separates storage, rights, page context, alt text, and reviewer approval.

공개 전 workflow

Collect the current bucket, object list, local photo folder, page URL, photographer memo, license email, current alt text, and publication plan. Mask private notes and do not publish images with people before review.

Create one registry row per image. The row should contain objectName, pageUrl, rights, credit, expiry, altText, publicScope, and reviewer. This is easier to review than a loose folder full of images.

Claude Code 범위와 사람 검토

Claude Code can create naming rules, custom metadata suggestions, alt text drafts, and missing-field questions. It can also mark unknown rights or visible people as review-needed.

Humans confirm rights, credits, people in the photo, public scope, deletion requests, and brand fit. If the rights field is blank, the image should not be published.

세 가지 Use case

Use case 1: objectName과 연결된 사진 대장

• Input: Cloud Storage object list, image folder, location, season, page URL.
• Output: objectName, place, season, use, page URL, replacement candidate.
• Human review: location match, old facility photos, people in image, forbidden use.

Use case 2: 권리와 공개 범위 검토

• Input: photographer memo, license email, contract terms, target page, social media plan.
• Output: web allowed, ads allowed, social blocked, expiry, credit required, review-needed.
• Human review: license scope, people, credit, expiration date, deletion contact.

Use case 3: 페이지 문맥에 맞춘 alt 문구

• Input: image, heading, nearby text, facility name, season, location.
• Output: alt text, caption draft, image replacement list.
• Human review: keyword stuffing, inaccurate location, invented content, brand wording.

복사해서 쓰는 프롬프트

Act as a tourism image management reviewer.
Goal: create a Cloud Storage image registry with rights, public scope, alt text, page URL, and human review.
Inputs: bucket, object list, page URL, location, photographer, license note, current alt text, people in photo.
Outputs: registry columns, naming rule, metadata suggestion, alt text drafts, review-needed list, publishing checklist.
Rules: never mark unknown rights as public, avoid keyword stuffing, keep credit and expiry visible, use /training/ as CTA.

첫 작업은 주요 관광 페이지에서 이미지 20장을 고르고 objectName, pageUrl, rights, credit, altText, publicScope, reviewer를 적는 것입니다. 권리 칸을 사람이 확인하기 전에는 공개 가능으로 표시하지 않습니다.

실행 가능한 확인 코드

const imageRecord = {
  bucket: "tourism-assets-prod",
  objectName: "kyoto/spring/kiyomizu-terrace-001.jpg",
  pageUrl: "/travel/kyoto/spring/",
  rights: "web use approved until 2027-03-31",
  creatorCredit: "Masa Photo",
  altText: "Spring view from Kiyomizu temple terrace in Kyoto",
  publicScope: "approved website pages only",
  reviewer: "content owner checks rights and people in photo",
  cta: "/training/"
};

const required = [
  "bucket",
  "objectName",
  "pageUrl",
  "rights",
  "creatorCredit",
  "altText",
  "publicScope",
  "reviewer",
  "cta"
];

const missing = required.filter((key) => !imageRecord[key]);

if (missing.length > 0) {
  throw new Error("Missing tourism image fields: " + missing.join(", "));
}

if (imageRecord.altText.length < 20 || imageRecord.altText.length > 125) {
  throw new Error("Alt text should be short but descriptive.");
}

if (!imageRecord.cta.startsWith("/")) {
  throw new Error("CTA must be an internal path.");
}

console.log("Tourism Cloud Storage image record is ready for rights review.");

The code checks whether the image registry has the minimum fields. A real implementation can run the same idea against exported object metadata, CMS image lists, or Markdown frontmatter.

Pitfall: 권리 확인 전 공개

Cause: the team assumes bucket storage equals publishing approval. Fix: keep a separate rights field and reviewer. Cause: alt text becomes a keyword list. Fix: describe the actual image and its page context.

Cause: credit and expiry live only in a page draft. Fix: keep them in the registry or metadata plan so they travel with the image.

자주 묻는 질문

Q. Is an image public just because it is in Cloud Storage?

A. No. Storage location and publishing approval are different. Rights, public scope, credit, and people in the photo need review.

Q. What should alt text contain?

A. It should describe what is visible and why it relates to the page. Avoid keyword stuffing and invented details.

Q. Where should the conversion path go?

A. For image governance and publishing workflow, send readers to training. The reader often needs a review process, not only a storage snippet.

상담 경로

The business path is training or consultation. A tourism site with many photos needs a registry, rights workflow, alt text template, and publishing approval path. The article shows the first twenty-image step.

실제로 확인한 결과

I verified the slug, frontmatter, internal CTA, external official references, executable JavaScript, code fence, FAQ, pitfall section, and human review path. The next action is to pick twenty live images and check blank rights or alt text fields.