Meta-Cognition Framework

Name: meta-reasoning-framework
Rating: 92
Author: codgician

Don't answer directly. Trace through cognitive layers first.

The Three Layers

code

L3: Content (WHY)       → What does the reader need?
    │
    ▼
L2: Architecture (WHAT) → How should it be structured?
    │
    ▼
L1: Implementation (HOW) → How do we code it?

Trace Directions

Entry Point	Direction	Example
L1 error	UP ↑	Build fails → Design issue? → Content mismatch?
L2 design	Both ↕	Check constraints (↑) → Implement (↓)
L3 feature	DOWN ↓	Feature → Structure → Code

The 3-Strike Rule

After 3 failed attempts at same layer → STOP → Trace to adjacent layer.

code

Strike 1: Try direct fix
Strike 2: Try alternative approach  
Strike 3: STOP - You're at wrong layer
         └─→ Trace UP or DOWN

5-Question Reboot (When Stuck)

#	Question	Purpose
1	What am I trying to achieve?	Goal clarity
2	What layer is this?	Layer identification
3	What constraints apply?	L3 check
4	What have I tried?	Pattern recognition
5	Should I trace UP or DOWN?	Direction decision

Concrete Examples

Example 1: Build Error (Entry: L1)

code

User: "Build fails with E0382 ownership error in Site.hs"

Trace UP:
├── L1: What's the immediate error? → Move after borrow
│       ↑
├── L2: Why is this data passed around? → Shared across functions
│       ↑
└── L3: What's the data's role? → Post metadata, immutable

Decision: Use reference, not ownership transfer

Example 2: New Feature (Entry: L3)

code

User: "Add dark mode support"

Trace DOWN:
├── L3: What does reader need? → System preference, manual toggle
│       ↓
├── L2: How to structure? → CSS variables, prefers-color-scheme
│       ↓
└── L1: How to implement? → SCSS variables, JS toggle, localStorage

Decision: CSS-first with JS enhancement

Example 3: Design Question (Entry: L2)

code

User: "Where should pagination logic live?"

Trace BOTH:
├── L3 ↑: What content patterns need pagination? → Posts, slides
├── L2: Current structure? → Posts has it, slides doesn't
└── L1 ↓: How to generalize? → Extract to Paginate.hs

Decision: Shared module with configurable URLs

Layer-Skill Mapping

Layer	Focus	Primary Skills
L3	Why (Content)	`skill-content-strategy`, `meta-reflection`
L2	What (Structure)	`fact-hakyll-architecture`, `skill-tdd-workflow`
L1	How (Code)	`skill-coding-standard`, `meta-verification-loop`

Cross-Layer Keywords

Keywords	Likely Layers	Load Skills
bilingual + error	L3 + L1	`skill-content-strategy` + `skill-coding-standard`
template + feature	L2 + L3	`fact-hakyll-architecture` + `skill-content-strategy`
refactor + design	L1 + L2	`skill-coding-standard` + `fact-hakyll-architecture`

Integration with Agents

code

┌─────────────────────────────────────────┐
│              User Input                  │
└─────────────────────────────────────────┘
                    │
                    ▼
┌─────────────────────────────────────────┐
│         Router (Auto-triggered)          │
│  • Detect keywords (400+ patterns)       │
│  • Identify entry layer                  │
│  • Load appropriate skills               │
└─────────────────────────────────────────┘
                    │
        ┌───────────┼───────────┐
        ▼           ▼           ▼
   ┌────────┐  ┌────────┐  ┌────────┐
   │ L3     │  │ L2     │  │ L1     │
   │ Planner│  │Tech-   │  │ Coder  │
   │        │  │Lead    │  │        │
   └────────┘  └────────┘  └────────┘
        │           │           │
        └───────────┼───────────┘
                    ▼
┌─────────────────────────────────────────┐
│            Validator                     │
│  • Evidence-based verification           │
│  • Cross-layer consistency check         │
│  • Final verdict                         │
└─────────────────────────────────────────┘
                    │
                    ▼
┌─────────────────────────────────────────┐
│        Continuous Learning               │
│  • Extract patterns from session         │
│  • Update skills if needed               │
│  • Persist memory for next session       │
└─────────────────────────────────────────┘