vaibhav-deveree/skill-awesome/universal-enterprise-multi-agent-code-review-production-readiness-skill

Universal Enterprise Multi-Agent Code Review & Production Readiness Skill

Purpose

This skill is designed for ANY software project:

• SaaS Platforms
• Web Applications
• Mobile Applications
• Java/Spring Boot Services
• Microservices
• APIs
• AI/LLM Products
• Internal Enterprise Systems
• FinTech Applications
• Healthcare Applications
• E-Commerce Platforms

The objective is to identify realistic production risks while minimizing false positives through multiple independent review layers and a final verification board.

Guiding Principles

1. Do not suggest feature additions.
2. Do not suggest redesigns unless required to fix a validated production issue.
3. Do not report style-only issues.
4. Do not report theoretical problems.
5. Every finding must be verified.
6. Every finding must survive a final review board.
7. Findings without evidence are rejected.

PHASE 0 – Project Discovery

Before reviewing code, perform discovery.

Identify:

• Architecture
• Languages
• Frameworks
• Databases
• Queues
• Caches
• Deployment model
• Authentication model
• Authorization model
• External integrations
• Monitoring stack
• Logging stack
• Build system

Produce:

Architecture Summary

Technology Stack Summary

Risk Surface Map

PHASE 1 – Multi-Agent Deep Review

Spawn independent review agents.

Each agent works separately.

Agents must not see findings from other agents.

Agent 1 – Frontend Engineer

Review:

• Components
• State Management
• Routing
• Rendering
• Forms
• Validation
• Accessibility
• Error Handling

Checks:

• Re-render loops
• Memory leaks
• State corruption
• UI race conditions
• Loading state failures

Agent 2 – Backend Engineer

Review:

• Controllers
• Services
• Business Logic
• DTOs
• Mappers
• APIs

Checks:

• Logic bugs
• Validation gaps
• Contract mismatches
• Concurrency issues

Agent 3 – Database Engineer

Review:

• Queries
• Indexes
• Transactions
• Locks

Checks:

• N+1 queries
• Missing indexes
• Deadlocks
• Full table scans

Agent 4 – Performance Engineer

Review:

• Latency
• Throughput
• Resource Usage

Checks:

• Hot paths
• Expensive operations
• Scaling bottlenecks

Agent 5 – Security Engineer

Review:

• Authentication
• Authorization
• Secrets
• Encryption

Checks:

• Access control issues
• Token handling
• Data leakage

Agent 6 – Reliability Engineer

Review:

• Failure handling
• Retry logic
• Recovery logic

Checks:

• Unhandled exceptions
• Service degradation paths
• Recovery failures

Agent 7 – QA Engineer

Review:

• User journeys
• Regression risks
• Edge cases

Checks:

• Broken flows
• Data inconsistencies
• Missing validation

Agent 8 – Integration Engineer

Review:

• External APIs
• Webhooks
• Third-party services

Checks:

• Retry handling
• Rate limiting
• Mapping consistency

PHASE 2 – Line-by-Line Inspection

Every file must be reviewed.

Inspect:

• Methods
• Classes
• Components
• Services
• Repositories
• Utilities

Look for:

• Null pointer risks
• Undefined access
• Race conditions
• Deadlocks
• Memory leaks
• Resource leaks
• Async issues
• Validation gaps

PHASE 3 – Cross File Validation

Review interactions between:

• APIs
• DTOs
• Services
• Events
• Queues
• Shared libraries

Detect:

• Contract mismatches
• Circular dependencies
• Serialization issues
• Hidden coupling

PHASE 4 – Worst Case Scenario Simulation

Simulate:

Infrastructure Failures

• Database outage
• Cache outage
• Queue outage
• API outage

User Behavior

• Double click
• Refresh during save
• Multiple tabs
• Concurrent users

Data Problems

• Null values
• Corrupted values
• Duplicate values
• Partial responses

Network Problems

• Timeout
• Packet loss
• Slow connections

PHASE 5 – Finding Verification Layer

For EVERY finding:

Spawn a Verification Agent.

Responsibilities:

1. Verify file location.
2. Verify line number.
3. Verify code path.
4. Verify reproducibility.
5. Verify production impact.
6. Verify issue is not already handled.

Verification Result:

• VALIDATED
• REJECTED

Rejected findings are removed.

PHASE 6 – Impact Analysis Layer

For every VALIDATED finding:

Spawn:

Business Impact Agent

Evaluates:

• Revenue impact
• Customer impact
• Operational impact

Production Impact Agent

Evaluates:

• Crash risk
• Data corruption risk
• Availability impact

Scalability Agent

Evaluates:

• Impact at 10 users
• Impact at 100 users
• Impact at 1000 users
• Impact at 10000 users

PHASE 7 – Final 4-Agent Review Board

This is mandatory.

Every finding must be reviewed by ALL four agents.

Board Agent 1 – Principal Architect

Validates:

• Technical correctness
• Root cause accuracy
• Fix feasibility

Decision:

• Approve
• Reject

Board Agent 2 – Principal QA

Validates:

• Reproducibility
• User impact
• Testability

Decision:

• Approve
• Reject

Board Agent 3 – Principal Reliability Engineer

Validates:

• Production risk
• Recovery implications
• Stability concerns

Decision:

• Approve
• Reject

Board Agent 4 – Engineering Manager

Validates:

• Priority
• Business value
• Implementation cost

Decision:

• Approve
• Reject

Final Verdict Rules

A finding is included ONLY IF:

• At least 3 of 4 board members approve.

Otherwise:

STATUS = REJECTED

Reason must be documented.

Finding Classification

Every approved finding must be classified as:

Confirmed Production Bug

Must fix.

Production Risk

Should fix.

Technical Debt

Can be deferred.

Monitoring Recommendation

Track but do not change.

False Positive

Discard.

Scoring Framework

Each finding receives:

Required Finding Format

Priority:

• P0
• P1
• P2
• P3

Status:

Classification:

File:

Line Number:

Method:

Category:

Description:

Failure Scenario:

Root Cause:

Impact:

Recommended Fix:

Verification Result:

Board Votes:

Confidence Score:

Implementation Plan

Generate:

Phase 1 – P0

Release blockers.

Phase 2 – P1

High-risk production issues.

Phase 3 – P2

Medium-risk issues.

Phase 4 – P3

Low-risk issues.

For each issue:

• Effort
• Risk
• Dependencies
• Rollout Strategy

Final Executive Report

Include:

Architecture Summary

Review Coverage

Validated Findings

Rejected Findings

False Positives

Technical Debt

Security Risks

Reliability Risks

Performance Risks

Scalability Risks

Integration Risks

Database Risks

Production Readiness Score (0-100)

Go / No-Go Recommendation

Final recommendation must be based ONLY on validated findings approved by the Final 4-Agent Review Board.