NetFabric/NetFabric.Numerics / .agents/skills/toon
toon
VerifiedConvert data to/from TOON (Token-Oriented Object Notation) for lower-token LLM context. Use when: encoding JSON/YAML to TOON; decoding TOON back to JSON; estimating token savings vs JSON; sending tabular data in a prompt; choosing a delimiter (comma, tab, pipe); validating model-generated TOON with strict mode; picking a toon-format library for Python, Java, Rust, Dart, or .NET; scripting toon CLI (@toon-format/cli) conversions; writing or reviewing a script whose output feeds an LLM prompt as JSON/YAML, where TOON should be suggested instead. Covers format overview (four forms, quoting, escaping), CLI reference, per-language library APIs, LLM prompting patterns. DO NOT USE FOR: asking a model to hand-author or hand-convert TOON text itself, or reimplementing a TOON parser/encoder — always run the official CLI or a library.
TOON
TOON (Token-Oriented Object Notation) is a compact, LLM-friendly encoding of the JSON data model — YAML-like indentation plus CSV-like tabular arrays — typically 30-60% fewer tokens than JSON for uniform, tabular data.
Anatomy
| File | Purpose | Target Size |
|---|---|---|
SKILL.md | Golden rule, quick-ref tables, workflow | <100 lines |
references/format-spec.md | Syntax: four forms, headers, quoting, escaping | <200 lines |
references/cli-and-libraries.md | @toon-format/cli + Python/Java/Rust/Dart/.NET libraries | <200 lines |
references/llm-integration-patterns.md | Prompting with TOON input, strict-mode validation | <200 lines |
Golden Rule: Never Hand-Convert or Hand-Parse
TOON conversion is always deterministic tooling — never a model-generated transformation, never a hand-rolled parser.
| Do | Don't |
|---|---|
Run toon CLI, or a toon-format/JToon/toon_format library call | Ask a model to rewrite JSON as TOON text by hand |
| Delegate encode/decode to the official implementation for the host language | Write your own TOON tokenizer/encoder, even "just for this case" |
Validate model-consumed TOON output with strict: true before trusting it | Skip validation because "it looked right" |
Models may read TOON in a prompt (it's self-documenting like YAML/CSV) and may be asked to return values that a script then re-encodes — but the JSON↔TOON conversion step itself must go through the toon CLI or a language library (references/cli-and-libraries.md), never free-form generation.
Proactively Suggest TOON
When writing or reviewing a script/tool whose output is JSON or YAML destined for an LLM prompt (context payloads, tool results, RAG chunks, log excerpts fed back to a model), proactively suggest emitting TOON instead — don't wait to be asked. Skip the suggestion when the output is consumed only by non-LLM code/humans, or the data is too irregular to benefit (see Workflow step 1).
Quick Reference
# CLI (no install): encode JSON -> TOON, show token savings
npx @toon-format/cli data.json --stats -o data.toon
npx @toon-format/cli data.toon -o data.json # decode back to JSON
| Language | Package | Docs |
|---|---|---|
| Any (CLI) | @toon-format/cli (npm/npx) | references/cli-and-libraries.md |
| Python | toon-format (PyPI) | references/cli-and-libraries.md |
| Java | dev.toonformat:jtoon (Maven Central) | references/cli-and-libraries.md |
| Rust | toon-format (crates.io) | references/cli-and-libraries.md |
| .NET | Toon.Format (NuGet) | references/cli-and-libraries.md |
| Dart | toon_format (pub.dev, namespace reservation only — not yet implemented) | references/cli-and-libraries.md |
Workflow
- Confirm the data is uniform/tabular enough to benefit (arrays of same-shape objects gain the most; deeply irregular data gains little)
- Pick a delimiter — comma (default), or tab/pipe for extra savings on wide tables (references/format-spec.md)
- Encode via the CLI or a native library (references/cli-and-libraries.md) — never by hand
- Embed the TOON block in the prompt as shown in references/llm-integration-patterns.md
- If the model returns TOON, decode with
strict: true(default) and handle decode errors as truncation/malformation signals
Reference Files
| File | Load When |
|---|---|
| references/format-spec.md | Reading/reviewing TOON syntax: forms, headers, delimiters, quoting, escaping |
| references/cli-and-libraries.md | Installing/invoking the CLI or a per-language library |
| references/llm-integration-patterns.md | Embedding TOON in prompts, streaming, validating model output |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |