本文へ移動
cccskills
無料GitHub で公開

wps-doc-scraper

Faithfully archive public WPS/KDocs/金山文档 links, especially embedded ProcessOn .pof mind maps and canvases, as raw source data, original SVG/PNG, and Markdown. Use when a user gives a kdocs.cn or wps.processon.com link and asks to scrape, save, download, 扒下来, 归档, or 转 Markdown without logging in or saving the document to an account.

インストール方法を見る

含まれるファイル(8)

  • SKILL.md4.3 KB
  • agents/openai.yaml228 B
  • references/capture-manifest.md1.2 KB
  • references/permission-and-failure-boundaries.md1.2 KB
  • references/processon-mindmap-api.md2.0 KB
  • references/rendered-svg-capture.md1.8 KB
  • scripts/render_svg_tiles.py6.8 KB
  • scripts/wps_processon_extract.py20.8 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

WPS Doc Scraper

Overview

Archive WPS/KDocs public documents with source fidelity. Prefer unauthenticated data APIs, keep the raw payloads, capture the original visual artifact when the document is a canvas or mind map, and generate Markdown only as a structured representation of the source.

This is an extraction skill, not a writing skill. Do not use an LLM to rewrite, summarize, smooth, or infer missing document content unless the user explicitly asks for a separate analysis after the archive is complete.

Workflow Decision Tree

  1. Identify the URL type.

    • kdocs.cn/view/l/<share_id> or kdocs.cn/l/<share_id>: read public link metadata first.
    • wps.processon.com/diagrams/view?...: extract file_id and group_id, then use the ProcessOn data API.
    • wps.processon.com/wpsapi/diagrams/view/api?...: treat as the source API directly.
    • Other WPS/KDocs pages: collect public metadata, try official public export/download only when available, then use browser DOM capture as a fallback.
  2. For ProcessOn .pof mind maps or canvases, run the API extractor:

    python3 /path/to/wps-doc-scraper/scripts/wps_processon_extract.py \
      --url "https://www.kdocs.cn/view/l/..." \
      --output-dir "/path/to/archive-dir"
    

    Expected outputs: processon-api.json, processon-definition.json, capture-manifest.json, and <title>.md.

  3. Capture the original image.

    • If the page exposes a rendered SVG, save the serialized full SVG as <title>-全画布.svg.
    • If the SVG uses foreignObject, do not trust ImageMagick alone for text rendering. Use scripts/render_svg_tiles.py to make the PNG through macOS Quick Look square tiles.
    python3 /path/to/wps-doc-scraper/scripts/render_svg_tiles.py \
      --svg "/path/to/<title>-全画布.svg" \
      --output "/path/to/<title>-全画布.png"
    
  4. Validate the archive.

    • Markdown is a structural conversion, not an editorial rewrite.
    • The original image exists for visual documents.
    • Raw JSON and manifest exist.
    • No output text contains the Unicode replacement character �.
    • Any failed or permission-limited path is recorded in the manifest or final response.

Hard Rules

  • Never log in, save to an account, copy into the user's cloud drive, or mutate the remote document unless the user explicitly requests it.
  • Never bypass CAPTCHAs, paywalls, tenant restrictions, or permission walls. A public link that requires login is a hard boundary unless the user provides authorized access.
  • Do not infer hidden nodes or missing text. Preserve what the API or rendered DOM actually exposes.
  • HTTP 200 is not enough. Detect login-wall payloads such as 用户未登录, empty definitions, and placeholder shells.
  • Keep raw source artifacts before transformation: API JSON, parsed definition JSON, serialized SVG, screenshots or PNGs, and a capture manifest.
  • For browser fallback, capture DOM/source data before screenshots when possible; screenshots alone are not a faithful text archive.
  • Report extraction gaps plainly. Do not silently produce a polished Markdown file from partial data.

Bundled Resources

scripts/

  • wps_processon_extract.py: deterministic extractor for public WPS/KDocs ProcessOn .pof mind maps. It resolves KDocs share metadata, downloads the ProcessOn data API JSON, parses the embedded definition, and writes Markdown plus a manifest.
  • render_svg_tiles.py: macOS Quick Look based SVG-to-PNG renderer for full-canvas SVGs with foreignObject text. It renders square vertical tiles and stitches them into one PNG.

references/

  • processon-mindmap-api.md: endpoint pattern and payload shape for WPS-hosted ProcessOn files.
  • rendered-svg-capture.md: browser-side process for extracting the rendered full-canvas SVG and producing a PNG.
  • capture-manifest.md: minimum manifest fields and acceptance checks.
  • permission-and-failure-boundaries.md: login walls, forbidden escalation, and failure reporting rules.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Fixes web search on an agent whose model backend can't run it: a relay/reseller proxying Claude or Codex returns empty instead of failing. Use when web search returns nothing, a model insists a shipped product doesn't exist, someone wants to give an agent internet access, or the user is on a third-party base URL, relay, or 中转站. Diagnoses which built-in tools are dead, removes them, and installs a working replacement.

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

抓取 A 股消息面情报:从财联社、华尔街见闻、金十、新浪 7x24、东财快讯、 证监会/央行/上交所/财政部政策公告、东方财富股吧等公开来源抓取与股票相关的 新闻、政策、情绪,输出结构化 JSON 或 Markdown。 当用户提到“A 股消息面”、“抓新闻”、“个股消息”、“政策监管”、“股吧情绪”、 “财联社”、“东财快讯”、“市场情绪”或需要把某只股票相关的公开情报聚合出来时 触发。也适用于“帮我看看 000001 最近有什么消息”这类口语化请求。

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

Transcribes audio or video to speaker-labeled, timestamped text, locally with MLX on Apple Silicon or remotely. Use for 转录 / 录音转文字 / 说话人分离 / 字幕, and also for preparing audio for ASR without transcribing: 转格式, 降采样到 16kHz, merging recorder segments, or compressing and speeding up audio before 飞书妙记 — even when it looks like a one-line ffmpeg job.

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

Routes audio: StepFun ASR/语音识别, StepFun TTS/配音, transcript/妙记→会议纪要, merge/review minutes. Reads one bundled specialist; generic ASR and correction stay direct.

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

Diagnoses and repairs repository setup and guarded Git workflows for Claude Code or Codex — environment repair, startup sync, hook auditing, collaborator handoff. Use when a repo won't run, a teammate onboards, hook output duplicates, or commit/push/conflict needs guarding. Not for lost-commit recovery (use git-safety-net), GitHub ops (use github-ops), or history scrubbing (use github-sensitive-data-cleanup).

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

Runs adversarial due-diligence on a benchmark the user envies — a founder, KOL, company, or product whose success looks inflated — splitting marketing bubble from real signal, then mapping the validated playbook onto the user's own resources. Use for 尽调/对标/拆解 a competitor, 抄/偷师 their playbook, or suspecting 水分/泡沫 in claims. Prefer over deep-research when debunking inflated claims, not a neutral briefing.

日本語の概要は準備中です。原文の説明を表示しています。

daymade/claude-code-skills1,4522026年10月11日 更新

daymade のスキルをすべて見る

このスキルの問題を報告する