360 Secure SkillHub
360 安全技能中心

QuiverAI Skill

这是基于QuiverAI的AI矢量图工具,核心功能分两类:一是文本生成SVG,可通过提示词+风格指导生成多变体矢量图,支持指定模型、调整创意度;二是栅格图像转SVG,支持本地/在线PNG/JPG/WebP转SVG,带自动裁剪等优化。典型场景包括创建图标、logo、插画,或矢量化现有图片,输出为本地SVG文件,使用前需配置API key。

lolieatapple
lolieatapple
热度 37.62
最近更新 2026-05-11 17:39:12
分类:创作依赖:代码执行 / 读写文件 / 第三方 API应用场景:图像生成 / 品牌素材 / 文件转换

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 quiver-ai-skill
Metadata
name
quiver
description
Generate SVGs from text prompts and convert raster images (PNG/JPG/WebP) to SVG using QuiverAI's AI models. Use when user asks to create icons, illustrations, logos as SVG, or to vectorize/convert an image to SVG.
argument-hint
<generate|vectorize|models> [options]
allowed-tools
Bash(quiver:*), Bash(pwd), Read, Write, Glob

QuiverAI SVG Generation Skill

You have access to the quiver CLI tool for AI-powered SVG generation and image vectorization.

Prerequisites

The quiver CLI must be installed globally:

npm install -g quiver-ai-cli

If the user hasn't configured their API key yet, guide them:

quiver config set api_key <their-key>

Commands

Text to SVG — quiver generate

Generate SVG graphics from a text description.

quiver generate [options] "<prompt>"

Options:

Flag Description
-m, --model <model> Model ID (default: arrow-preview)
-o, --output <file.svg> Save to file instead of stdout
-s, --stream Stream via SSE (shows progress on stderr)
-n, --count <n> Number of SVG variants (default: 1)
-t, --temperature <float> Creativity level (default: 1)
-i, --instructions <text> Additional style guidance
--max-tokens <n> Upper bound on output tokens
-r, --reference <url|file> Reference image (repeatable)

Image to SVG — quiver vectorize

Convert a raster image into a clean SVG vector.

quiver vectorize [options] <image-path-or-url>

Options:

Flag Description
-m, --model <model> Model ID (default: arrow-preview)
-o, --output <file.svg> Save to file instead of stdout
-s, --stream Stream via SSE
-t, --temperature <float> Temperature (default: 1)
--auto-crop Auto-crop to dominant subject
--target-size <px> Resize before processing
--max-tokens <n> Upper bound on output tokens

List Models — quiver models

quiver models list
quiver models get <model-id>

Configuration — quiver config

quiver config set api_key <key>
quiver config set default_model <model>
quiver config get
quiver config path

Workflow

When the user asks you to create or generate an SVG:

  1. Run pwd to confirm the working directory.
  2. Determine whether this is a generate (text-to-SVG) or vectorize (image-to-SVG) task.
  3. Always use -o <filename>.svg to save the output to a file so the user can use it.
  4. Pick a sensible filename based on the prompt content (e.g., rocket-icon.svg).
  5. Run the quiver command.
  6. Read the generated SVG file to verify it was created successfully.
  7. Report the result: file path, file size, and a brief summary.

Arguments Handling

If the user provides arguments directly after /quiver, pass them through:

quiver $ARGUMENTS

Examples

User says "create a rocket icon":

quiver generate -o rocket-icon.svg "a minimalist rocket icon, clean lines, flat design"

User says "make me 3 variants of a logo":

quiver generate -n 3 -o logo.svg "a modern tech startup logo with geometric shapes"

This produces logo-1.svg, logo-2.svg, logo-3.svg when n > 1.

User says "convert this image to SVG" with a file path:

quiver vectorize -o vectorized.svg --auto-crop ./photo.png

User says "vectorize this URL":

quiver vectorize -o output.svg https://example.com/image.png

User says "list available models":

quiver models list

Tips

  • Always save output to a file with -o so the user has a usable artifact.
  • Use --auto-crop for vectorize when the source image has whitespace or background.
  • For icons and simple graphics, a lower temperature (e.g., -t 0.7) gives more consistent results.
  • For creative/artistic SVGs, a higher temperature (e.g., -t 1.2) adds variety.
  • Use -i instructions to control style: "flat design", "line art", "monochrome", "gradient", etc.
  • When generating multiple variants, mention which files were created.
  • If quiver is not found, tell the user to install it: npm install -g quiver-ai-cli
  • If auth fails, tell the user to run: quiver config set api_key <key>