Skip to main content

AI Agent Skill

The JsonCut Agent Skill is a curated set of markdown files that teach AI coding agents how to use the JsonCut API for image and video generation. It works with any agent that supports skill/instruction files.

GitHub Repository

Supported Agents​

AgentSetup
Claude CodeAdd to .claude/settings.json or reference CLAUDE.md
OpenAI CodexReference AGENTS.md as project instruction
OpenClaw / CustomPoint agent at AGENTS.md as entry point

What's Included​

The skill provides structured documentation that agents can reference to build correct API requests:

  • API Reference — Authentication, endpoints, job polling, error handling, and token costs
  • Image Generation — All 6 layer types (image, text, rectangle, circle, gradient, html), positioning, and effects
  • Video Generation — 16+ layer types, 65 transitions, clips, audio, subtitles, and globalLayers
  • HTML Layers — HTML/CSS rendering with TailwindCSS, GSAP, Chart.js, D3.js, Lottie, and more
  • Ready-to-use Examples — Complete image and video configurations with cURL, Python, and Node.js

Quick Setup​

Claude Code​

Add the skill as a custom instruction source:

{
  "permissions": {},
  "customInstructions": "Read @jsoncut-skill/AGENTS.md for JsonCut API usage."
}

Codex / Other Agents​

Point your agent to the AGENTS.md file in the repository. It serves as the main entry point and links to all detailed documentation.

How It Works​

The agent reads the structured markdown files to understand:

  1. How to authenticate with the JsonCut API (x-api-key header)
  2. How to build image and video configurations as JSON
  3. How to create jobs, poll for completion, and retrieve results
  4. What layer types, properties, and transitions are available
  5. Token costs and rate limits

This enables AI agents to generate correct API calls without trial and error.