Google Docs·Notion·Slack·GitHub·webpage clipboard HTML sample corpus 수집

[developer-1px]

배경

#19 조사에서 current editor는 text/html paste importer를 지원하지 않는다고 정리했다. Rich HTML paste를 제품 범위로 올리려면 sanitizer 구현보다 먼저 실제 외부 앱 clipboard HTML corpus가 필요하다.

근거

• docs/editor-paste-html-security-policy-audit.md
• ProseMirror changelog: pasted HTML parser, Trusted Types, OS clipboard comments, table/wrapper/isolating node fixes 반복
• Lexical changelog: Google Docs paste, HTML conversion, style import/export, table paste, script ignore, URL sanitize 반복

수집 대상

각 sample은 raw MIME payload와 editor 기대 결과를 같이 남긴다.

• Google Docs: heading, bold/italic/link, list, table-ish content
• Notion: paragraph, heading, callout/code/list, link
• Slack: message, mention-like text, link, emoji, inline code/code block
• GitHub rendered page: markdown-rendered heading/list/code/link/table
• Generic webpage/article: paragraph/link/image/style/script wrapper

기록할 것

• text/html
• text/plain
• text/markdown이 있으면 함께
• text/uri-list가 있으면 함께
• source app/browser/OS
• selection shape: partial inline, whole block, multi-block
• current expected: HTML ignored, plain/markdown fallback only
• future importer expected: allowed model nodes and dropped attributes

완료 기준

• 5개 source app의 raw clipboard MIME sample이 저장되어 있다.
• 각 sample에 sanitizer/drop expectation이 있다.
• unsafe URL/style/class/event-handler fixture가 최소 하나 이상 있다.
• current reader expectation과 future HTML importer expectation을 분리한다.
• sample corpus가 붙여넣기/드롭의 HTML·plain text·uri-list 파싱 실패 조사 #10/pasted HTML sanitization·Trusted Types·보안 경계 조사 #19 후속 HTML importer 설계의 test fixture로 연결된다.

[developer-1px]

현재 agent 환경에서는 이 이슈의 완료 기준을 충족할 수 없습니다.

이 이슈는 실제 Google Docs, Notion, Slack, GitHub rendered page, generic webpage/article에서 OS/browser clipboard에 올라오는 raw MIME payload corpus를 수집해야 합니다. 이 데이터는 로그인 상태, source app editor, OS clipboard, browser clipboard implementation에 의존하므로 로컬 저장소 코드 변경이나 synthetic clipboard fixture로 대체할 수 없습니다.

필요한 다음 입력:

• 5개 source app별 raw text/html, text/plain, 가능한 경우 text/markdown, text/uri-list
• source app/browser/OS, selection shape
• current expectation: HTML ignored, plain/markdown fallback only
• future importer expectation: allowed model nodes and dropped attributes
• unsafe URL/style/class/event-handler sample 최소 1개

corpus가 제공되면 #10/#19 후속 HTML importer 설계의 fixture로 연결할 수 있습니다.

[developer-1px]

Update after PR #126:

Added a repo-tracked collection contract for this issue.

• Manifest: tests/fixtures/clipboard-html-corpus/manifest.json
• Sample convention: tests/fixtures/clipboard-html-corpus/README.md
• Contract doc: docs/editor-clipboard-html-corpus-collection.md
• CI guard: packages/editable/clipboardCorpus.test.ts

Current status: still collecting. This issue should stay open until raw clipboard payload samples are added for Google Docs, Notion, Slack, GitHub rendered pages, and a generic webpage/article.

[developer-1px]

Update after PR #127:

Added two real browser clipboard samples to the corpus:

• github-rendered/heading-list-code-link-table.json: copied from a GitHub-rendered Markdown document; includes text/html and text/plain clipboard MIME payloads.
• generic-webpage/paragraph-link-image.json: copied from a generic article page; includes image/link/table content plus raw class/style-heavy browser clipboard HTML.

Also tightened packages/editable/clipboardCorpus.test.ts so listed samples must match declared source/shape entries and contain non-empty required MIME payloads.

Current remaining #74 source/shape gaps:

• Google Docs: heading-marks-link, list, table-ish
• Notion: paragraph-heading, callout-code-list, link
• Slack: message-mention-link-emoji, inline-code-code-block
• Generic webpage: unsafe-style-script-wrapper

Refs PR #127.

[developer-1px]

Update after PR #128:

Expanded the clipboard corpus with additional real browser clipboard samples:

• Google Docs: heading-marks-link, list, table-ish
• Notion: one public-page sample covering paragraph-heading, callout-code-list, and link
• Generic webpage: unsafe-style-script-wrapper showing Chromium strips inline handlers, javascript: URLs, and scripts before clipboard HTML while style/class remain

Current #74 coverage:

• Google Docs: complete
• Notion: complete
• GitHub rendered page: complete
• Generic webpage/article: complete
• Slack: still missing message-mention-link-emoji and inline-code-code-block

Remaining blocker for closing #74 is authenticated Slack web/app clipboard data. Refs PR #128.

[developer-1px]

Update after PR #129:

Added a standalone capture page for the remaining clipboard samples:

• Tool: tools/evidence/clipboard-capture.html
• Usage doc: tools/evidence/README.md
• Evidence doc: docs/editor-evidence-capture-tools.md

Remaining #74 Slack fixture targets:

• tests/fixtures/clipboard-html-corpus/slack/message-mention-link-emoji.json
• tests/fixtures/clipboard-html-corpus/slack/inline-code-code-block.json

After capture, add both files to tests/fixtures/clipboard-html-corpus/manifest.json. Refs PR #129.