Skill: Toon

Owner: Lythaeon

Summary: Token-saving utilities for agents: Read JSON, list directories, search code, and fetch APIs in the compact TOON format. Reduces token usage by 30-50% for structured data.

Tags: latest:3.0.0

Version history:

v3.0.0 | 2026-02-01T17:11:49.927Z | user

Summary: Major update removes built-in CLI scripts; shifts usage to global TOON CLI.

• Removed internal utility scripts: bin/toon.js and lib/commands.js.
• Updated documentation: Agents must now install and use the official global toon CLI.
• Instructions shifted from custom skill commands to standard Unix piping with global toon.
• Emphasized token savings, speed, and prompt brevity by preferring global CLI.

v1.0.1 | 2026-02-01T16:24:30.200Z | user

• Removed 4 files, including all tests and utility scripts.
• Updated API fetching: replaced fetch command with more flexible curl, supporting all curl arguments.
• Adjusted documentation and examples to use curl instead of fetch for API calls.

v1.0.0 | 2026-02-01T16:02:36.352Z | user

Initial release of toon-utils skill.

• Introduces token-saving utilities for agents by using the compact TOON format.
• Supports reading JSON, listing directories, searching code, and fetching APIs with reduced token usage (30-50% savings for structured data).
• Provides clear command guidelines to prioritize TOON-based operations over traditional methods.
• Adds examples for correct and incorrect usage.
• Implementation leverages the @toon-format/toon library for encoding via Node.js.

Archive index:

Archive v3.0.0: 2 files, 1216 bytes

Files: SKILL.md (1651b), _meta.json (129b)

File v3.0.0:SKILL.md

TOON (Token-Oriented Object Notation) Skill

This skill provides instructions for reducing token usage by 30-60% when dealing with structured data (JSON, directory structures, API responses).

Setup & Installation

To minimize command overhead and save tokens, you MUST install the official TOON CLI globally before usage:

# Check if toon is installed, install if missing
toon --version || npm install -g @toon-format/cli

Guidelines for Agents

• Always optimize JSON: Convert JSON output to TOON before including it in your context.
• Minimize Command Length: Use the direct toon command instead of npx to keep your own prompt usage low.
• Prefer pipes: Use standard Unix pipes to send data directly to the toon converter.

Commands & Usage

1. Reading JSON Files

Instead of reading raw JSON, pipe it to toon.

cat data.json | toon

2. Fetching APIs

Pipe external API responses directly into toon.

curl -s https://api.example.com/data | toon

Supports all curl flags (e.g., -X POST, -H "Authorization: ...").

3. Listing Directory Structures

Use tree -J or any JSON-outputting tool and pipe to toon.

tree -J path/to/dir | toon

4. Converting In-Line Data

To compress a JSON string for your context:

echo '{"key":"value"}' | toon

Why install TOON?

• Command Token Savings: toon is shorter than npx @toon-format/cli, saving tokens every time you run a command.
• Execution Speed: Local installation is significantly faster than on-demand fetching.
• Readability: TOON is designed to be highly readable for LLMs.

File v3.0.0:_meta.json

{ "ownerId": "kn7400d5fktsezh1662yyq7xz180bgxs", "slug": "toon-utils", "version": "3.0.0", "publishedAt": 1769965909927 }

Archive v1.0.1: 4 files, 3231 bytes

Files: bin/toon.js (1178b), lib/commands.js (4094b), SKILL.md (1985b), _meta.json (129b)

File v1.0.1:SKILL.md

description: >- Token-saving utilities for agents: Read JSON, list directories, search code, and fetch APIs in the compact TOON format. Reduces token usage by 30-50% for structured data. metadata: openclaw: emoji: "📉" requires: anyBins: ["node"] keywords: ["tokens", "optimization", "json", "utils", "search"] homepage: "https://github.com/openclaw/skill-toon"

TOON Skill

Provides utilities to generate TOON (Text Only Object Notation) output. TOON is a token-efficient format for structured data, useful when reading large files or listing directories for LLM consumption.

Guidelines for Agent

You MUST prioritize using this skill for structured data.

• Reading JSON: ALWAYS use skills/toon/bin/toon.js cat <file> instead of read_file for .json or .rc (json) files. This saves significant tokens.
• Listing Directories: ALWAYS use skills/toon/bin/toon.js ls <dir> instead of ls -R or find. (Generates a directory tree object).
• Fetching API: ALWAYS use skills/toon/bin/toon.js curl <args...> when fetching JSON APIs. Supports all curl arguments (e.g. -X POST).
• Searching Code: ALWAYS use skills/toon/bin/toon.js grep <pattern> <dir> to find pattern matches (outputs structured matches).
• Converting JSON: Use skills/toon/bin/toon.js convert '<json_string>' to convert raw JSON strings to TOON format.

Examples

Correct (Save Tokens):

skills/toon/bin/toon.js cat src/config.json
skills/toon/bin/toon.js curl https://api.github.com/repos/openclaw/openclaw/issues
skills/toon/bin/toon.js curl -X POST -H "Content-Type: application/json" -d '{"foo":"bar"}' https://api.example.com/data
skills/toon/bin/toon.js grep "TODO" src/agents

Incorrect (Wastes Tokens):

cat src/config.json # DO NOT DO THIS for JSON files

Implementation

This skill uses the @toon-format/toon library to perform encoding. It is a lightweight wrapper that runs via node.

File v1.0.1:_meta.json

{ "ownerId": "kn7400d5fktsezh1662yyq7xz180bgxs", "slug": "toon-utils", "version": "1.0.1", "publishedAt": 1769963070200 }