agentsclimarketplace

Case 03460

Skill knownasnaffy/prompthound/dataset/case_03460

A fast, offline static risk analysis CLI for AI agent skill files. Detects malicious instructions, steganographic payloads, and dangerous capability chains.

Install
npx -y skills add knownasnaffy/prompthound --skill case_03460

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate SVG images using text LLM instead of image generation APIs. Use when user wants to create illustrations, icons, cartoons, diagrams, or any visual content without DALL-E/Midjourney/Stable Diffusion. The LLM writes SVG code directly - works great for simple graphics, cartoons, icons, and stylized art.

SKILL.md

6.1 KB, as published. Nobody here has run it

SVG Artist

Generate images by writing SVG code with the text model. No image generation API needed.

How It Works

  1. User describes the image → "画一只可爱的小猫" / "create a cartoon cat"
  2. LLM writes SVG code → Calculates coordinates, shapes, colors
  3. Convert to PNG → Using rsvg-convert or convert

Quick Start

User: 给我画一只小狗
You: [Generate SVG code] → Send the PNG

SVG Drawing Guide

Canvas Setup

Default size: 400x400 pixels. Origin (0,0) is top-left.

<svg width='400' height='400' xmlns='http://www.w3.org/2000/svg'>
  <!-- Background -->
  <rect width='400' height='400' fill='#87CEEB'/>
  <!-- Your shapes here -->
</svg>

Basic Shapes

Circle

<circle cx='200' cy='200' r='50' fill='#FF6B6B' stroke='#333' stroke-width='2'/>

Ellipse (great for bodies, heads)

<ellipse cx='200' cy='200' rx='80' ry='50' fill='#D2691E' stroke='#8B4513' stroke-width='3'/>

Rectangle (legs, buildings)

<rect x='100' y='300' width='50' height='80' rx='10' fill='#666'/>

Polygon (ears, stars)

<polygon points='150,50 130,100 170,100' fill='#8B4513'/>

Path (curves, tails, mouths)

<!-- Arc -->
<path d='M 100 200 A 50 50 0 0 1 200 200' stroke='#333' stroke-width='3' fill='none'/>
<!-- Bezier curve -->
<path d='M 100 200 Q 150 150 200 200' stroke='#333' stroke-width='3' fill='none'/>

Proportional Design Tips

Use percentages of canvas size (400px):

  • Head: rx='70' ry='60' (about 15-20% of width)
  • Body: rx='100' ry='70' (about 25% of width)
  • Eyes: r='15' (about 4% of width)

Position using fractions:

  • Center X: 200 (50%)
  • Head Y: 140 (35%)
  • Body Y: 250 (62%)

Color Palette

Warm colors: #D2691E (chocolate), #FF6B6B (coral), #FFB347 (orange) Cool colors: #87CEEB (sky blue), #98FB98 (pale green), #DDA0DD (plum) Neutrals: #8B4513 (saddle brown), #808080 (gray), #F5F5DC (beige)

Common Patterns

Eyes (cute style)

<!-- Eye white -->
<ellipse cx='180' cy='130' rx='15' ry='15' fill='white' stroke='black' stroke-width='2'/>
<!-- Pupil -->
<ellipse cx='183' cy='132' rx='7' ry='7' fill='black'/>
<!-- Highlight -->
<ellipse cx='186' cy='130' rx='3' ry='3' fill='white'/>

Blush/cheeks

<ellipse cx='150' cy='160' rx='15' ry='10' fill='#FFB6C1' opacity='0.6'/>

Smile

<path d='M 180 170 A 25 25 0 0 0 220 170' stroke='#8B4513' stroke-width='3' fill='none'/>

Tongue

<ellipse cx='200' cy='190' rx='10' ry='15' fill='#FF69B4'/>

Step-by-Step Workflow

When user asks for an image:

Step 1: Understand the Subject

  • What animal/object? → dog, cat, house, car, etc.
  • What style? → cute, realistic, minimalist, cartoon
  • What pose? → sitting, standing, happy, sad

Step 2: Plan the Composition

Head position: cy = 35% (140)
Body position: cy = 62% (250)
Legs: cy = 78% (310)
Background: soft gradient or solid color

Step 3: Write SVG Code

Use Node.js inline to generate:

node -e "
const fs = require('fs');
const svg = \`<svg>...</svg>\`;
fs.writeFileSync('/tmp/image.svg', svg);
"

Step 4: Convert to PNG

rsvg-convert /tmp/image.svg -o /tmp/image.png
# or: convert /tmp/image.svg /tmp/image.png

Step 5: Send the Image

<qqimg>/tmp/image.png</qqimg>

Example: Drawing a Cat

<svg width='400' height='400' xmlns='http://www.w3.org/2000/svg'>
  <rect width='400' height='400' fill='#E6E6FA'/>

  <!-- Body -->
  <ellipse cx='200' cy='250' rx='100' ry='70' fill='#FFB347' stroke='#FF8C00' stroke-width='3'/>

  <!-- Head -->
  <ellipse cx='200' cy='140' rx='70' ry='60' fill='#FFB347' stroke='#FF8C00' stroke-width='3'/>

  <!-- Ears -->
  <polygon points='130,110 110,50 160,90' fill='#FF8C00'/>
  <polygon points='270,110 290,50 240,90' fill='#FF8C00'/>

  <!-- Eyes -->
  <ellipse cx='170' cy='130' rx='15' ry='20' fill='#90EE90' stroke='black' stroke-width='2'/>
  <ellipse cx='230' cy='130' rx='15' ry='20' fill='#90EE90' stroke='black' stroke-width='2'/>

  <!-- Pupils (vertical slits) -->
  <ellipse cx='170' cy='130' rx='4' ry='12' fill='black'/>
  <ellipse cx='230' cy='130' rx='4' ry='12' fill='black'/>

  <!-- Nose -->
  <ellipse cx='200' cy='160' rx='8' ry='5' fill='#FF69B4'/>

  <!-- Whiskers -->
  <line x1='120' y1='155' x2='165' y2='160' stroke='#666' stroke-width='1.5'/>
  <line x1='120' y1='165' x2='165' y2='165' stroke='#666' stroke-width='1.5'/>
  <line x1='235' y1='160' x2='280' y2='155' stroke='#666' stroke-width='1.5'/>
  <line x1='235' y1='165' x2='280' y2='165' stroke='#666' stroke-width='1.5'/>

  <!-- Mouth -->
  <path d='M 190 170 Q 200 180 210 170' stroke='#FF8C00' stroke-width='2' fill='none'/>

  <!-- Legs -->
  <ellipse cx='150' cy='310' rx='20' ry='30' fill='#FFB347' stroke='#FF8C00' stroke-width='2'/>
  <ellipse cx='250' cy='310' rx='20' ry='30' fill='#FFB347' stroke='#FF8C00' stroke-width='2'/>

  <!-- Tail -->
  <path d='M 300 250 Q 340 200 320 150' stroke='#FFB347' stroke-width='20' fill='none' stroke-linecap='round'/>
</svg>

Script: generate_svg.js

Use the helper script for common subjects:

node scripts/generate_svg.js "cute puppy with big eyes" /tmp/puppy.png

The script provides templates and the LLM fills in details.

Limitations

  • Best for: cartoons, icons, simple illustrations, stylized art
  • Not ideal for: photorealistic images, complex scenes, detailed textures
  • Works well with: animals, characters, objects, simple backgrounds

Tips for Better Results

  1. Start simple - Basic shapes first, add details after
  2. Use layers - Background → body → head → features → highlights
  3. Add character - Big eyes, blush, expressive features
  4. Keep proportions - Cute style = big head, small body
  5. Test incrementally - Generate, check, adjust

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.