Triple Order

Adaptive Layered Reasoning for AI

Triple Order is a platform-independent AI skill framework for organizing complex knowledge, problems, and systems. It uses Domain → Branch → Concept as a default cognitive scaffold, while adapting to flat, two-layer, extended, graph, and hybrid structures when the task requires it.

Triple Order is not a rule that everything has three layers. It is a method for finding the right order of abstraction.

Other languages: 简体中文

---

Why Triple Order?

Human knowledge is often organized in three visible layers:

Philosophy   →  Nietzsche's Philosophy  →  Will to Power
Software Eng →  Software Process        →  Waterfall Model
Psychology   →  Child Psychology        →  Oedipus Complex

This works because three layers can answer three questions at once:

1. What domain does this belong to?
2. What branch or category within that domain?
3. What is the specific object, model, or concept?

But not every problem fits three layers. Some need only two. Some need five. Some are networks, not trees.

Triple Order gives AI agents a disciplined way to:

• Start from a three-layer default;
• Detect when three layers are wrong;
• Switch to flat, extended, graph, or hybrid structures;
• Preserve cross-cutting relationships that a pure tree would lose;
• Explain why it chose a particular structure.

---

What's Inside

triple-order/
├── SKILL.md              # Core behavior specification (read by AI agents)
├── manifest.yaml         # Machine-readable metadata
├── schemas/              # Input/output JSON schemas
├── examples/             # Positive, negative, and edge cases
├── evals/                # Test cases and scoring rubric
├── adapters/             # Platform-specific loading notes
└── LICENSE

---

Quick Start

For AI Agents

Load SKILL.md into the agent's system context. The skill will:

1. Analyze the user's task (classification, explanation, learning, design, decision).
2. Determine the appropriate structural representation.
3. Emit a structure in Markdown or JSON, along with the reason for the choice.

For Developers

The skill is executable-free and network-free. It consists of prompts, schemas, and examples only.

# manifest.yaml (excerpt)
name: triple-order
compatibility:
  - claude
  - openai
  - gemini
  - generic-agent
security:
  executable: false
  external_network: false

Adapter notes for specific platforms live in adapters/.

---

When to Use

Use Triple Order when the AI needs to:

• Organize a body of knowledge for learning or documentation;
• Decompose a complex product, system, or design problem;
• Produce a structured explanation with clear context;
• Choose between a list, a tree, or a knowledge graph representation;
• Explain relationships between multiple domains that partially overlap.

When Not to Use

Do not invoke Triple Order for:

• Trivial rewrites, translations, or single-sentence answers;
• Simple lookups where a flat list is sufficient;
• Tasks with no structural component (e.g. formatting a phone number).

The skill's own SKILL.md includes negative rules to prevent it from forcing three layers onto everything.

---

Core Principles

1. Three layers is a default scaffold, not a universal law.
2. Distinguish classification, causation, composition, and influence.
3. Depth follows task and knowledge scale, not aesthetics.
4. When a tree cannot express reality, switch to a graph or hybrid.
5. Always explain why a structure was chosen.
6. Ship reasoning skills with negative examples and evaluations.

---

Status

Version: 0.1.0 (MVP) License: Apache-2.0 (code) / CC BY 4.0 (documentation and prompts)

Contributions, examples, and evaluation cases are welcome.