toon-format
TOON (Token-Oriented Object Notation) encoding for LLM-efficient data representation. 30-60% token savings vs JSON for structured data.
What this skill does
# TOON Format Skill
TOON is a compact, human-readable encoding of JSON designed for LLM prompts.
## Variables
| Variable | Default | Description |
|----------|---------|-------------|
| VALIDATION_MODE | strict | `strict` (CLI validation required) or `lenient` (parser permissive) |
| AUTO_FIX | true | Run fix scripts automatically on validation failure |
## Instructions
**MANDATORY** - Follow the format rules in this skill and `reference/format-rules.md`.
- Always include explicit `[count]` for arrays
- Use 2-space indentation for tabular rows
- Add blank line after tabular arrays to terminate them
- Use semicolons for nested arrays in cells (NOT commas)
## Red Flags - STOP and Reconsider
If you're about to:
- Put commas inside quoted strings in tabular cells
- Omit the `[count]` from array declarations
- Forget the blank line after a tabular array
- Use YAML-style `- item` list syntax
- Add comments with `#`
**STOP** -> Check `reference/format-rules.md` -> Then proceed
## Workflow
1. [ ] Determine if TOON is appropriate (see "When to Use TOON" below)
2. [ ] Design data structure for tabular representation
3. [ ] **CHECKPOINT**: Verify format matches rules
4. [ ] Write TOON with proper indentation
5. [ ] Validate with `npx @toon-format/cli --decode`
6. [ ] If errors, run auto-fix scripts from `reference/validation-tools.md`
## Cookbook
### Format Rules
- IF: Writing or editing TOON files
- THEN: Read `reference/format-rules.md`
- COVERS: Syntax rules, constraints, correct patterns
### Validation & Fixing
- IF: TOON file fails validation
- THEN: Read `reference/validation-tools.md`
- COVERS: CLI validation, auto-fix scripts
## When to Use TOON
**Good candidates:**
- Uniform tabular data (lists of similar objects)
- Documentation indexes and manifests
- Repeated key-value structures
- Any structured data being fed to LLMs
**Use JSON instead for:**
- Deeply nested structures
- Non-uniform data (mixed schemas)
- Data requiring frequent programmatic manipulation
- Configuration files read by tools
## TOON Syntax
### Key-Value (YAML-like)
```toon
name: John
age: 30
active: true
```
### Arrays with Explicit Length
```toon
tags[3]: red, green, blue
```
### Tabular Data (Primary Use Case)
```toon
# Syntax: name[count]{col1,col2,col3}:
# Followed by 2-space indented CSV-like rows
users[3]{id,name,email}:
1,John,[email protected]
2,Jane,[email protected]
3,Bob,[email protected]
```
**IMPORTANT**: Rows MUST be indented with 2 spaces. Always add a blank line after the last row.
### Nested Arrays in Cells
Use semicolons or quote values containing multiple items:
```toon
pages[2]{path,keywords,priority}:
/intro,"overview;basics;start",core
/api,"reference;methods;types",core
```
### Multi-line Strings
Use quoted strings with `\n` for line breaks:
```toon
description: "This is a multi-line\nstring value that preserves\nline breaks."
```
### Nested Objects
```toon
config:
database:
host: localhost
port: 5432
cache:
enabled: true
```
## Encoding Patterns for Documentation
### Index Files
```toon
meta:
category: libraries
generated: 2025-01-15T10:30:00Z
count: 5
items[5]{id,name,description,path,keywords}:
baml,BAML,Structured LLM outputs,ai-docs/libraries/baml/_index.toon,"llm;types;structured"
mcp,MCP,Tool integration protocol,ai-docs/libraries/mcp/_index.toon,"tools;context;servers"
prisma,Prisma,Type-safe database ORM,ai-docs/libraries/prisma/_index.toon,"database;orm;sql"
```
### Page Summaries
```toon
meta:
library: baml
page: error-handling
source_url: https://docs.boundaryml.com/guide/error-handling
content_hash: a3f2c1d4e5
summary:
purpose: "Configure retry policies and fallback strategies for resilient LLM calls."
key_concepts[4]: RetryPolicy,FallbackClient,timeout,exponential backoff
gotchas[2]: Default timeout 60s may be too short,Retries count against rate limits
code_patterns[1]:
- lang: baml
desc: Retry policy configuration
code: "retry_policy Resilient {\n max_retries 3\n strategy exponential\n}"
```
## Token Comparison
**JSON** (~280 tokens):
```json
{"pages":[{"path":"/intro","section":"guide","title":"Introduction","priority":"core"},{"path":"/setup","section":"guide","title":"Setup","priority":"core"}]}
```
**TOON** (~100 tokens):
```toon
pages[2]{path,section,title,priority}:
/intro,guide,Introduction,core
/setup,guide,Setup,core
```
**Savings: ~64%**
## Tools & Libraries
- **TypeScript/JS**: `@toon-format/toon` (npm)
- **Python**: `python-toon` (pip)
- **Online**: https://toontools.vercel.app/playground
- **Spec**: https://toonformat.dev/
## Best Practices
1. Design data structures for tabular representation when possible
2. Use explicit `[count]` for arrays - helps LLM parsing
3. Keep JSON as canonical source, TOON for LLM transport
4. Use semicolons for nested arrays (e.g., `"val1;val2;val3"`)
5. Truncate long strings (code snippets < 200 chars)
6. Always add a blank line after tabular arrays to terminate them
## Format Constraints
**Enforced by official `@toon-format/cli` (use `--decode` to validate):**
| Rule | ❌ Incorrect | ✅ Correct |
|------|-------------|-----------|
| No comments | `# comment` | (remove comments entirely) |
| No YAML-style lists | `- item` | `key[N]{col}:` + indented rows |
| Rows must be indented | `row,data` | ` row,data` (2 spaces) |
| Arrays need count+cols | `items{a,b}:` | `items[3]{a,b}:` |
| No multiline strings | `key: \|` | `key: "line1\nline2"` |
| No pipe delimiters | `val\|val` | `val,val` or `"val;val"` |
| No commas in tabular cells | `"a,b"` | `"a;b"` |
| Blank line AFTER arrays | missing terminator | blank line after last row |
**Validation tools:**
```bash
# Validate with official CLI (use --decode, not validate!)
npx @toon-format/cli --decode file.toon > /dev/null
# Auto-fix common issues (run in this order)
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_comments.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_yaml_lists.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_nested_lists.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_commas.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_pipes.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_multiline.py path/
python3 .claude/ai-dev-kit/dev-tools/scripts/fix_toon_blank_lines.py path/
```
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.