本文へスキップ
TakoAPI
スキル一覧に戻る

visual-prompt-engine

abdullah4ai 作

Coding Agents & IDEs
380ダウンロード数はclawskills.shが公開している値です
0閲覧数
0件のいいねサインインするとこのスキルにいいねできます
—まだ評価がありません
あなたの評価

概要

Generate diverse, non-repetitive image prompts powered by real visual references from Dribbble and design platforms.

説明

このスキルでできること

Generates image prompts grounded in real Dribbble design references rather than internal AI patterns. A scraping pipeline converts trending shots into style cards with palette, composition, mood, and lighting data. Prompt history is tracked to prevent the same structures from repeating.

Grounds prompts in real, trending design work rather than the model's internal pattern library, which tends to recycle the same visual vocabulary.

使いどころ

  • Getting a fresh image prompt without recycling the same visual clichés
  • Generating UI illustration prompts based on current design trends
  • Creating diverse prompts for a batch of product renders
  • Building mood board descriptions from trending Dribbble references
  • Exploring visual directions for a brand identity concept

ワークフロー例

入力

User asks: Give me a unique image prompt for a minimal tech product shot

エージェント
  1. 1

    Read style_cards.json to find cards matching a minimal, tech-oriented aesthetic

  2. 2

    Select 1-2 style cards and extract palette, composition, and lighting values

  3. 3

    Check prompt_history.json to identify structures used in the last 5 prompts

  4. 4

    Compose a new prompt using an unused pattern from prompt-patterns.md combined with the style card data

  5. 5

    Append the prompt to history and return it

出力

A specific image prompt with hex color references, a named composition structure, lighting direction, and mood terms drawn from a real Dribbble shot

要件

  • Browser tool such as Playwright or Camofox (recommended for Dribbble collection)

インストール

clawhub install abdullah4ai/visual-prompt-engine

CLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。

リンク