AgentSkillsCN

cot-research

采用深度思维链研究方法,系统性地分析问题。借鉴行业领先企业的前沿模式(如Canvas、Moodle、Coursera等LMS平台;Google、Meta、Netflix等架构领域的佼佼者)。适用于调试复杂问题、规划重大重构,或分析系统设计难题的场景。

SKILL.md
--- frontmatter
name: cot-research
description: Deep Chain-of-Thought research methodology for systematic problem analysis. Applies SOTA patterns from industry leaders (Canvas, Moodle, Coursera for LMS; Google, Meta, Netflix for architecture). Use when debugging complex issues, planning major refactors, or analyzing system design problems.

CoT Research Skill

Systematic Chain-of-Thought (CoT) research methodology for deep analysis of technical problems, following SOTA patterns from industry leaders and academic research (as of January 2026).

When to Use This Skill

  • Debugging complex, multi-layer issues
  • Planning major refactors or architecture changes
  • Analyzing system design problems
  • Ensuring solutions follow industry best practices
  • Avoiding patch-fix approaches in favor of root cause solutions

Core Methodology

Step 1: Problem Definition

Clearly articulate the problem before diving into solutions:

  • What is the observed symptom?
  • What is the expected behavior?
  • When did the problem first appear?

Step 2: Root Cause Analysis (5 Whys)

Apply iterative questioning to find the deepest source:

code
Step 1: Why is [symptom] happening?
  → Because [immediate cause]

Step 2: Why is [immediate cause] happening?
  → Because [deeper cause]

Step 3: Why is [deeper cause] happening?
  → Because [even deeper cause]

... Continue until you reach a root cause that is actionable

Step 3: Logic Flow Mapping

Identify all related logic flows:

FlowComponentsConnections
Data FlowInput → Processing → OutputDatabases, APIs
Control FlowUser actions → System responsesEvents, Handlers
Error FlowExceptions → Recovery → LoggingTry/Catch, Fallbacks

Step 4: SOTA Research

Research current best practices from:

For LMS/EdTech:

  • Canvas LMS (Instructure)
  • Moodle (open source)
  • Coursera, edX, Udemy
  • Google Classroom

For General Architecture:

  • Google Cloud Architecture Center
  • AWS Well-Architected Framework
  • Microsoft Azure Architecture Guide
  • Netflix Tech Blog
  • Meta Engineering

For Database/SQL:

  • PostgreSQL official docs
  • Use-The-Index-Luke.com
  • PGAnalyze best practices

Step 5: Comparison Analysis

Compare current implementation vs SOTA:

AspectCurrentSOTA PatternGap
Design.........
Performance.........
Maintainability.........

Step 6: Solution Design

Create solution that:

  1. Addresses root cause (not symptoms)
  2. Follows SOTA patterns
  3. Is maintainable long-term
  4. Avoids patch-fix approaches

Output Format

Quick Analysis (For Simple Issues)

markdown
## Problem
[One-line description]

## Root Cause
[2-3 sentences explaining why]

## Solution
[Specific actionable fix]

Deep Analysis (For Complex Issues)

markdown
## 🔍 Deep Analysis Report

### 1. Problem Statement
[Clear description of symptom and expected behavior]

### 2. Root Cause Chain (CoT)
Step 1: Why? → [answer]
Step 2: Why? → [answer]
Step 3: Why? → [answer]
...
Root Cause: [final answer]

### 3. Related Logic Flows
[Diagram or table of data/control flows]

### 4. SOTA Comparison
| Aspect | Current | Industry Best | Gap |
|--------|---------|---------------|-----|
| ... | ... | ... | ... |

### 5. Proposed Solution
[Detailed approach following SOTA]

### 6. Implementation Plan
- [ ] Step 1: ...
- [ ] Step 2: ...
- [ ] Step 3: ...

Examples

Example 1: Type Mismatch Analysis

Problem: Hibernate validation fails on startup

CoT Analysis:

code
Step 1: Why is Hibernate failing?
  → Type mismatch: DB column is INTEGER, Entity field is BigDecimal

Step 2: Why is there a type mismatch?
  → Entity was created assuming NUMERIC for currency values

Step 3: Why does DB have INTEGER?
  → Original migration used INTEGER for simplicity

Step 4: Which is correct per SOTA?
  → For percentage scores (0-100): INTEGER is correct
  → For currency/precise decimals: NUMERIC is correct

Root Cause: Entity type doesn't match domain semantics
Solution: Change Entity type to Integer (for scores)
         OR change DB to NUMERIC (for currency)

Example 2: Duplicate Tables Analysis

Problem: Schema has multiple similar tables

CoT Analysis:

code
Step 1: Why are there duplicate tables?
  → enrollments, learning_enrollments exist with similar structure

Step 2: Why were both created?
  → Historical evolution: direct course enrollment → class-based

Step 3: Which pattern is SOTA?
  → Canvas/Moodle: Class-based enrollment (Section/Group)
  → Students enroll in CLASS, not directly in COURSE

Root Cause: Schema evolved without removing legacy tables
Solution: Keep class-based table, remove legacy

Constraints

  • DO NOT propose patch fixes that address symptoms only
  • DO NOT skip root cause analysis and jump to solutions
  • DO NOT ignore SOTA patterns from industry leaders
  • ALWAYS document the reasoning chain
  • ALWAYS compare against industry best practices
  • ALWAYS consider long-term maintainability

Research Resources (Jan 2026)

LMS Domain

PostgreSQL

Architecture Patterns

Trigger Phrases

This skill should be activated when user mentions:

  • "Suy nghĩ thật kỹ" (think carefully)
  • "think step-by-step"
  • "Chain of Thought" or "CoT"
  • "Root cause analysis"
  • "SOTA patterns"
  • "Best practices from large organizations"
  • "Không sửa chắp vá" (no patch fixes)