AI-Assisted Enterprise Full Stack Debugging

Purpose

This skill is designed for debugging complex production and full stack issues where:

• Issues are unclear or vaguely reported
• Multiple system layers are involved (UI, API, backend, DB)
• Production incidents or release blockers occur
• Root cause is not immediately obvious

When to Use

Use this skill for:

• Production incidents
• Regression issues
• Cross-service failures
• Performance degradation
• Intermittent or hard-to-reproduce bugs
• Customer-reported unclear issues

Core Principle

Never assume. Never guess. Always validate with evidence.

Only rely on:

• Logs
• Metrics
• Traces
• Code behavior
• Reproduction steps

Step 1: Understand the Problem

Convert vague issue into a clear technical problem statement and identify:

• Frontend
• Backend
• API
• Database
• Infrastructure

Step 2: Collect Full Context

Gather:

• UI behavior & screenshots
• API request/response logs
• Backend logs
• DB state
• Infra metrics (if any)

Step 3: Reproduce Issue

• Reproduce consistently
• Compare working vs failing cases
• Reduce to minimal steps

Step 4: AI Debugging Prompt

Use this in Claude Code / OpenCode:

You are a production-grade full stack debugging assistant.

Problem: {{issue_description}}

Context:

• UI: {{ui}}
• API: {{api}}
• Backend: {{backend}}
• DB: {{db}}
• Infra: {{infra}}
• Steps: {{steps}}

Tasks:

1. Convert into technical problem
2. Map full flow UI → API → backend → DB
3. Identify failure points
4. Correlate logs
5. Provide top root causes
6. Separate facts vs assumptions
7. Suggest fix + validation

Rules:

• No guessing
• Evidence-based only

Step 5: Root Cause Analysis

• Trace full flow
• Identify first failure point
• Validate using logs & metrics

Step 6: Fix Strategy

• Minimal safe fix
• No wide impact changes
• Validate all flows

Step 7: Validation & Prevention

• Verify fix in staging/prod-like env
• Ensure no regression
• Add monitoring if needed
• Document RCA

Output Format

Always return:

• Issue summary
• Root cause
• Evidence
• Fix
• Validation steps
• Prevention actions

Principle

Evidence > assumptions. Always.