salvador

"Have no fear of perfection - you'll never reach it"
-- Salvador Dalí

salvador is an autonomous visualization agent for Claude Code
                 it lays down the path from thinking about the Universe to looking at the Universe

• starts from a sketch
• inspects it with a headless browser
• critiques the aesthetics and UX
• and iteratively refines the result until the laws of physics hold and the visual matches the design principles

$ /visualize schrödinger's cat

how to play

install it

clone the repository:

git clone https://github.com/tolitius/salvador.git
cd salvador

start the agent:

claude

the Claude Code agent automatically detects the .claude configuration and loads the skill.

play

once salvador is in your .claude/skills path, start claude and visualize:

/visualize covalent bond H2O

or visualize the near cosmic forces:

/visualize the difference between throwing a basketball on Earth vs. on the Moon

or explain a concept:

/visualize the difference between a centralized network and a distributed mesh network

salvador will enter an autonomous loop: coding -> inspecting -> refining
when it is satisfied with the quality, it will launch the result in your browser

structure

the skill organization follows the proposed agent skills convention:

.
├── .claude/
│   ├── commands/
│   │   └── visualize.md       # slash command definition
│   ├── skills/
│   │   └── salvador/
│   │       ├── SKILL.md       # the brain: logic & constraints
│   │       ├── scripts/       # tooling: setup.sh, inspect.js
│   │       └── templates/     # scaffolding: index.html, main.js
│   └── settings.json          # permissions (auto-allows bash/edit)
└── CLAUDE.md                  # root agent context

license

Copyright © 2026 tolitius

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.