360 Secure SkillHub
360 安全技能中心

Dribbble + Pinterest Search

该 Skill 指导使用英文标签在 Dribbble 与 Pinterest 执行免登录公开搜索:先按主题选择平台与标签集,从搜索页收集候选卡片,必要时进入详情页补全高清封面、作者与发布时间,最终输出字段可为空的规范化记录,并说明选择器与公开时间戳不稳定。

joeyyyy0430
joeyyyy0430
热度 17.46
最近更新 2026-04-02 10:26:48
分类:搜索依赖:联网浏览搜索应用场景:研究 / 推荐

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add dribbble-pinterest-search
Metadata
name
dribbble-pinterest-search
description
Choose public search tags for Dribbble and Pinterest, run public search flows without login, and extract normalized result fields such as title, high-resolution image, author, source URL, and publish/update time when available. Use when Codex needs to collect design inspiration from Dribbble or Pinterest, design a keyword taxonomy for B-end/data-visualization themes, or explain how to get public search result fields from these two platforms.

Dribbble Pinterest Search

Use English tags, not Chinese keywords, for public design search on Dribbble and Pinterest. Treat search pages as candidate discovery and detail pages as field enrichment.

Keep the workflow public-only:

  • Do not require cookies, sessions, or login-only APIs.
  • Prefer browser automation for both platforms.
  • Use detail pages only when search results do not expose a reliable high-resolution image, author, or publish/update time.

Workflow

  1. Choose a category and tag set.

    • For B-end/data-visualization use cases, load references/tag-taxonomy.md.
    • Prefer Dribbble for dashboard, scene, motion, and high-quality UI shots.
    • Prefer Pinterest for color, infographic, isometric, and reference-heavy categories.
  2. Build the public search URL.

    • Dribbble: https://dribbble.com/search/<urlencoded-tag>
    • Pinterest: https://www.pinterest.com/search/pins/?q=<urlencoded-tag>&rs=typed
  3. Collect candidate cards from the search page.

    • Dribbble: find a[href*="/shots/"], then keep only links matching /shots/<numeric-id>.
    • Pinterest: find a[href*="/pin/"], then keep only links matching /pin/<numeric-id>/.
  4. Normalize search-card fields first.

    • Always capture at least: title, platform, source_url, cover_image_url, keyword/tag.
    • If search-card author or time is weak, leave it empty and enrich from the detail page.
  5. Enrich from the detail page when needed.

    • Use detail pages for high-resolution cover images.
    • Use detail pages for author_name.
    • Use detail pages for publish_time or updated_time only when the public page exposes it reliably.
  6. Return a normalized public-only record.

    • Keep unstable fields nullable.
    • Record which tag matched.
    • Do not fabricate publish times or authors.

Extraction Rules

Use these field priorities:

  • title: search-card text first, then detail-page meta/title.
  • cover_image_url: prefer highest srcset candidate or detail-page og:image.
  • author_name: prefer detail-page author metadata or title pattern.
  • publish_time: prefer detail-page og:updated_time, createdAt, datePublished, or equivalent. Leave null when unavailable.
  • summary: use normalized title or public description snippet.

Platform-specific details live in:

Guardrails

  • Stay within public pages and public metadata.
  • Expect DOM changes and anti-bot behavior; selectors are heuristics, not guarantees.
  • Treat Dribbble publish time as unstable on public pages.
  • Treat Pinterest og:updated_time as the best public timestamp unless richer relay data is needed.
  • Prefer English tags even for Chinese design topics.

Output Shape

Normalize results into a shape close to:

{
  "title": "",
  "platform": "dribbble|pinterest",
  "source_url": "",
  "cover_image_url": "",
  "author_name": "",
  "publish_time": null,
  "keyword": "",
  "category": "",
  "matched_tags": []
}