AgentSkillsCN

quiz-crawler

Velt Comments 在 React、Next.js 以及各类 Web 应用程序中的实现模式与最佳实践。当您需要为应用添加协作式评论功能、多种评论模式（自由式、弹出式、流式、文本式、页面式）、富文本编辑器评论（TipTap、SlateJS、Lexical）、媒体播放器评论，或图表评论时，可选用此方案。

SKILL.md
--- frontmatter
name: quiz-crawler
description: "Extract all questions/answer options from marketing quiz funnels and record final redirect/product URL + screenshots. Use when a user gives a quiz URL and asks to collect questions/answers or capture the offer page. Use output format specified in assets/output.md"

Workflow (reproducible pipeline)

This skill is for building a swipe database: store raw artifacts + a clean Notion page.

1) Crawl (deterministic path)

Run:

  • •node scripts/crawl.js "<quiz_url>"

Popup handling:

  • •Auto-accepts JS dialogs (alert/confirm/prompt).
  • •Attempts to click modal CTAs like Continue/Submit/OK when a popup blocks progress.

Optional:

  • •OUT_DIR=/path/to/output
  • •MAX_STEPS=60 (default 40)
  • •MAX_TIME_MS=420000 (default 300000)
  • •JPG_QUALITY=80
  • •FILL_FORMS=true|false (default true)

Artifacts written to OUT_DIR:

  • •results.json (steps + pickedAnswer + stopReason + finalUrl)
  • •step-XX.jpg (screenshot per interaction)
  • •step-XX.html (HTML snapshot when available)

2) Sync artifacts to R2 (deterministic)

Run:

  • •SRC_DIR=<OUT_DIR> node scripts/r2_sync.js

Required env:

  • •R2_ENDPOINT
  • •R2_BUCKET
  • •R2_PUBLIC_BASE (e.g. https://pub-....r2.dev)
  • •R2_ACCESS_KEY_ID
  • •R2_SECRET_ACCESS_KEY

Optional env:

  • •R2_PREFIX (default quiz)
  • •RUN_ID (otherwise derived from SRC_DIR)
  • •QUIZ_KEY (stable key for the funnel; otherwise derived from SRC_DIR)
  • •INCLUDE_HTML=true|false (default true)

Output:

  • •writes r2-manifest.json into SRC_DIR with public URLs.

Setup (minimal deps)

  • •

    OCR uses the system tesseract binary (no JS OCR deps).

  • •

    If tesseract is missing, install it:

    • •Ubuntu/Debian: sudo apt-get update && sudo apt-get install -y tesseract-ocr
    • •macOS (brew): brew install tesseract
  • •

    R2 sync requires these env vars. Recommended: store them in a skill-local .env file at ~/skills/quiz-crawler/.env (the scripts auto-load it).

    • •R2_ENDPOINT
    • •R2_BUCKET
    • •R2_PUBLIC_BASE
    • •R2_ACCESS_KEY_ID
    • •R2_SECRET_ACCESS_KEY
  • •

    Notion publish requires:

    • •NOTION_API_KEY
    • •NOTION_PARENT_PAGE_ID

.env notes:

  • •Lines like KEY=value or KEY="value".
  • •Do not commit .env (secrets). If you package/share the skill, exclude .env.

3) Extract Q/A via OCR (quick) + Publish Notion page

Run:

  • •SRC_DIR=<OUT_DIR> node scripts/ocr_extract.js (writes ocr.json)
  • •SRC_DIR=<OUT_DIR> node scripts/qa_extract.js (writes qa.json: OCR question + DOM options)
  • •SRC_DIR=<OUT_DIR> BRAND=<BrandName> NOTION_PARENT_PAGE_ID=<page_id> node scripts/publish_notion.js

Fast defaults:

  • •OCR is used only for the question (line containing ?).
  • •Options come from the crawler’s captured button text (more reliable than OCR).

Required env:

  • •NOTION_API_KEY
  • •NOTION_PARENT_PAGE_ID

Output:

  • •Creates 1 Notion child page under the parent with the format specified in assets/output.md:
    • •Question: ...
    • •Answer: ...
    • •Screenshot: <external image>

Data handling

  • •Use sensible dummy data to progress:
    • •name/email/phone/dates as needed.
  • •Date picker policy: always pick a random date when a date-picker is present.

Notes:

  • •This skill currently captures a single deterministic path. Different answers can lead to different pages; add branching only when requested.