AI Modernize Plugin
Codebase modernization assessment that finds technical debt, anti-patterns and quality issues in older AI-generated or legacy code.
Overview
Codebases built with older AI tools (Claude 3-era models, early Cursor, GPT-4 in 2024) or through "vibe coding" sessions can contain patterns that newer models avoid. This plugin finds those issues and produces a prioritized remediation roadmap with time estimates that assume AI-assisted work. It runs in Claude Code, Codex and Cursor.
Installation
Claude Code:
/plugin marketplace add charlesjones-dev/ai-agent-plugins
/plugin install ai-modernize@ai-agent-plugins
Codex:
codex plugin marketplace add charlesjones-dev/ai-agent-plugins
codex plugin add ai-modernize@ai-agent-plugins
Cursor: add this repository as a team marketplace, or copy this folder to ~/.cursor/plugins/local/ai-modernize. See the Cursor install steps.
Skills
/modernize-audit
Audits the whole repository, a directory or a single file, and saves a scored report with a phased roadmap.
It takes optional arguments after the command: a path for the scope, key=value settings that answer the setup questions, and --defaults. Quote values that contain spaces. In Codex and Cursor, type the same text after the skill name.
| Argument | Sets | Values |
|---|---|---|
| A path | Scope | A file or directory. . means the whole repository |
ai= | AI tool history | Tool names separated by commas, or unsure or none |
era= | When the code was written | A period such as 2024, or unsure |
stack= | Technology stack | The stack, or auto to use the detected one |
categories= | Assessment categories | all, or category numbers or names separated by commas, such as security,performance |
severity= | Severity threshold | all, medium (Medium and above) or high (High and Critical only) |
--defaults | Everything the other arguments leave out | No value. Skips the setup questions and uses the defaults |
Usage:
/modernize-audit
/modernize-audit src/api categories=security,performance
/modernize-audit . ai=copilot era=2024 stack=auto categories=all severity=medium
/modernize-audit src/services --defaults
/modernize-audit src/services/auth.ts --defaults severity=high
With no arguments, the skill asks about:
- AI tools and models used to build the codebase
- When the codebase was written
- Technology stack confirmation (auto-detected first)
- Which of the 12 assessment categories to run
- Audit scope (entire solution, or a specific file or directory)
- Severity threshold for the report
It asks only about what the arguments leave out, and asks nothing when all six are given. With --defaults, or in an unattended run, it asks nothing and uses defaults for anything missing: AI tools and era "not sure", the detected stack, all categories, the entire solution and all findings. When it runs all categories, it skips any that can't apply to the scope, such as Frontend Debt for backend-only code, and lists them as not applicable.
/modernize-scan is now /modernize-audit <path> --defaults (since version 2.0.0). It writes the full audit report rather than the shorter scan report.
Assessment Categories
| # | Category | What It Checks |
|---|---|---|
| 1 | SOLID/DRY/KISS Violations | God classes, duplicated logic, over-engineering, mixed paradigms, YAGNI |
| 2 | Type Safety & Language Misuse | any overuse, missing type guards, loose typing, language idiom violations |
| 3 | Error Handling | Empty catch blocks, swallowed errors, missing error boundaries, console.log debugging |
| 4 | Security Anti-patterns | Hardcoded secrets, missing validation, injection risks, insecure defaults |
| 5 | Performance Anti-patterns | N+1 queries, sync bottlenecks, missing pagination, full library imports |
| 6 | Testing Gaps | Implementation-coupled tests, over-mocking, missing edge cases, no integration tests |
| 7 | Architecture Debt | Tight coupling, circular deps, business logic in UI, missing abstraction layers |
| 8 | Frontend Debt | Prop drilling, state mismanagement, useEffect misuse, inline styles, missing a11y |
| 9 | Dependency Health | Deprecated packages, vulnerable versions, unnecessary imports, missing lock files |
| 10 | AI Hallucination Artifacts | Non-existent APIs, wrong signatures, hallucinated packages, deprecated methods |
| 11 | Modern Pattern Gaps | Missing modern syntax, outdated patterns, old CSS, legacy build tools |
| 12 | Configuration & DevOps Debt | Hardcoded config, missing env validation, no health checks, poor Docker practices |
Report Output
Reports are saved to docs/modernize/ in the repository, with timestamped filenames: YYYY-MM-DD-HHMMSS-modernize-audit.md.
Report Includes
- Modernization Score (0-100) with category breakdown
- Findings grouped by severity (Critical, High, Medium, Low)
- Exact file paths and line numbers for every finding
- Before/after code examples for remediation
- Time estimates that assume AI-assisted work
- Phased modernization roadmap with prioritized checklist
- A "Why Older AI Models Did This" note on each finding
Supported Technology Stacks
The plugin auto-detects the stack and adds language-specific checks for:
- JavaScript/TypeScript: React, Vue, Nuxt, Next.js, Angular, Svelte, Express, Fastify, Node.js
- C#/.NET: nullable reference types, async I/O, records, pattern matching, minimal APIs
- Python: type hints, f-strings, context managers, match statements, dataclasses
Plugin Details
| Field | Value |
|---|---|
| Version | 2.0.0 |
| Author | Charles Jones |
| License | MIT |
| Repository | ai-agent-plugins |
Contributing
Bug reports and fixes are welcome. See CONTRIBUTING.md.