SuperAPI AI Directory

ai-diagnostic

General & other

View repository

Updated 10-15 23:26

About this skill

ai-diagnostic

This is an AI-based multidimensional diagnostic analysis skill. Using the Supervisor pattern, it performs parallel diagnosis across 4 dimensions: technology, process, people, and business.

Overview

It provides a systematic diagnostic framework for identifying root causes beyond simple symptom-based resolution. It simultaneously analyzes 4 dimensions (technology, process, people, and business) and integrates the results to generate a comprehensive diagnostic report.

Core Features

  • Multidimensional Parallel Diagnosis: Simultaneous analysis across 4 dimensions: technology, process, people, and business
  • Root Cause Identification: Utilizes 5 Whys, Fishbone Diagram, and Iceberg Model techniques
  • Hypothesis-Based Diagnosis: Formulates verifiable hypotheses and assigns priorities
  • Actionable Solutions: Prioritized Action Items and ROI estimates

Trigger Keywords

  • diagnosis, diagnostic, problem analysis, troubleshooting, cause analysis
  • root cause, system diagnosis, status check, problem solving
  • performance degradation, error cause, bottleneck identification, issue analysis

Workflow

Step 1: Problem Collection

Systematically collect symptoms and context:

  • Symptom identification: error messages, performance metrics, user complaints
  • Impact assessment: number of affected users, business impact, urgency
  • Timeline: when the problem occurred, change trends, patterns
  • Environment information: system specifications, configuration, recent changes

Step 2: Hypothesis Generation

Generate verifiable hypotheses based on symptoms:

  • Hypotheses by dimension: possibilities by technology, process, people, and business dimensions
  • Prioritization: sorting based on likelihood and impact
  • Establish verification methods: define methodologies to verify each hypothesis

Step 3: Parallel Diagnosis

Diagnose 4 dimensions in parallel:

Dimension 1: Technology Diagnosis

  • Infrastructure: servers, network, database, third-party APIs
  • Code: bugs, performance bottlenecks, resource leaks, exception handling
  • Architecture: scalability, availability, complexity, technical debt
  • Security: vulnerabilities, permission management, data protection

Dimension 2: Process Diagnosis

  • Workflow: business processes, approval systems, handovers
  • Automation: whether repetitive tasks are automated, CI/CD pipelines
  • Monitoring: logs, alerts, response procedures
  • Documentation: Runbook, incident reports, knowledge base

Dimension 3: People Diagnosis

  • Competency: technology stack, domain knowledge, problem-solving ability
  • Communication: team collaboration, information sharing, decision-making
  • Workload: burnout, resource allocation, on-call burden
  • Culture: accountability, learning culture, tolerance for failure

Dimension 4: Business Diagnosis

  • KPI: performance metrics, SLA, goal achievement rate
  • Cost: infrastructure cost, development cost, opportunity cost
  • Customer: NPS, churn rate, complaints
  • Strategy: alignment between business goals and technical goals

Step 4: Identify Root Causes

Derive root causes using various techniques:

  • 5 Whys: Reach the root cause by repeating "Why?" 5 times
  • Fishbone Diagram: Categorize causes by category (Man, Machine, Material, Method, Environment)
  • Iceberg Model: Explore potential causes beneath the surface

Step 5: Propose Solutions

Present actionable solutions for each root cause:

  • Short-term actions: Immediately actionable mitigation measures (24-72 hours)
  • Mid-term solutions: Structural improvements to resolve root causes (1-4 weeks)
  • Long-term prevention: Systemic improvements to prevent recurrence (1-3 months)
  • Prioritization: Alignment based on ROI (return on investment) and urgency

Usage examples

Example 1: Diagnosing web service response time degradation

증상:
- 지난 2주간 API 평균 응답 시간 200ms → 800ms로 악화
- 오후 2시~4시에 심각 (최대 3초)
- 사용자 불만 50건/일 접수

컨텍스트:
- 2주 전 신규 기능 배포 (검색 엔진 변경)
- AWS, RDS MySQL 사용
- 온콜 엔지니어 1명, 번아웃 위험

Diagnosis result overview:

  • Technology: New search engine N+1 queries occur, DB CPU 95%, cache miss rate 45%
  • Process: Deployment without performance testing, no alerts set up, no response manual
  • People: 24/7 impossible with 1 on-call person, lack of knowledge transfer, lack of performance optimization capability
  • Business: Churn rate increased by 2.1%, SLA violations, new signup conversion rate dropped by 15%

Root cause: Absence of quality processes (performance testing, code review) and shortage of personnel acting in combination

Solutions:

  • [Urgent] Roll back new search engine, RDS scale-up, increase connection pool
  • [Mid-term] Optimize N+1 queries, build performance testing pipeline, increase on-call to 2 people
  • [Long-term] Establish code review culture, write Runbook, strengthen monitoring

Example 2: Diagnosing startup team productivity decline

증상:
- 분기별 기능 릴리스 5개 → 2개로 감소
- 스프린트 완료율 60% (이전 85%)
- 팀원 사기 저하, 퇴사 2명 발생

컨텍스트:
- 시리즈 A 투자 유치로 팀 10명 → 25명으로 급성장
- PO 3명, 각자 다른 우선순위
- 기술 부채 누적, 재택근무 전환

Diagnosis result overview:

  • Technology: Technical debt 40%, slow CI/CD, 2.5-day wait for code review
  • Process: Sprint scope keeps changing, daily meeting lengthened to 45 minutes
  • People: No onboarding program, mentoring time consumption, increased isolation
  • Business: Failure to manage investor expectations, slow hiring speed, 70% achievement of revenue target

Root cause: Delay in process, culture, and infrastructure upgrades due to rapid organizational growth

Solutions:

  • [Urgent] Establish single PO priority, freeze sprint scope, 15-minute daily meeting
  • [Mid-term] Onboarding program establishment, CI/CD optimization, code review SLA
  • [Long-term] Hiring an Engineering Manager, documentation culture, team building activities

Output format

The diagnostic report includes the following sections:

Required sections

  1. Symptom summary: A concise summary of observed symptoms, impact, and temporal patterns
  2. Hypothesis list: A list of verifiable hypotheses, priorities, and verification methods
  3. Diagnostic results by dimension: Results of the 4-dimension analysis, discovered issues and evidence
  4. Root cause analysis: Application of 5 Whys/Fishbone techniques, identification of root causes
  5. Solution priorities: Short/mid/long-term solutions, owners, expected duration, impact
  6. Execution plan: Timeline-based execution roadmap, milestones

Optional sections

  • Risk assessment: Risks of executing solutions and mitigation measures
  • Alternative comparison: Comparison of pros and cons of multiple solutions
  • Monitoring plan: Metrics and alerts for tracking success

Precautions

Precautions during diagnosis

  1. Distinguishing symptoms from causes: Clearly distinguish observed phenomena (symptoms) from root causes
  2. Bias prevention: Avoid confirmation bias and base conclusions on data
  3. Multidimensional thinking: Consider all 4 dimensions rather than focusing on only one
  4. Verifiability: Hypotheses must be formulated in a verifiable form
  5. Prioritization: Do not try to solve everything at once; approach in order of impact

Avoiding common mistakes

  • Treating only symptoms: Repeatedly applying temporary measures without resolving root causes
  • Single-dimension bias: Looking only at technical problems and ignoring process/people/business
  • Lack of data: Diagnosis relying only on intuition without evidence
  • Too many hypotheses: Limit hypotheses to an actionable level and assign priorities
  • Ignoring feasibility: Proposing ideal solutions without considering resources and capabilities

Related skills

  • moai-workflow-spec: When structural analysis of the problem is needed
  • moai-workflow-project: When understanding the project context is needed
  • moai-foundation-thinking: When a strategic thinking framework is needed

Post-processing chain (narrative deliverables such as diagnostic reports, root cause analysis, solution descriptions, etc.)

After writing sentence-based deliverables such as diagnostic reports (root cause analysis, solution descriptions), finish with the chain below.

  • moai-coworker:ai-slop-reviewer: Review and revise AI-sounding expressions, exaggeration, and clichés (required)
  • moai-writer:korean-humanize: Korean naturalness correction — after slop review (required)