🦨 code-smells

Stop shipping smelly code. 120+ named code smells with before/after examples across C, C++, C#, Java, Python, Go, Rust, JavaScript, TypeScript & SQL — an AI agent skill that improves code quality during both development and code review.

What It Does

Drop this skill into your AI coding agent and get an expert code reviewer that:

Identifies code smells by name — not just "this looks bad"
Explains why each smell is harmful and what it costs you
Shows concrete before/after refactoring examples
Prioritizes findings by severity so you fix the right things first
Covers 10 languages in one skill — no separate tools to configure

Languages

C · C++ · C# · Java · Python · Go · Rust · JavaScript · TypeScript · SQL

Smell Categories

Category	Examples
Naming & Linguistic	Abstract Names, Linguistic Confusion, Comedian Methods, Spelling Mistakes
Structure & Complexity	Arrow Code, Long Parameter List, Boolean Trap, Cyclomatic Complexity
OOP	God Objects, Shotgun Surgery, Primitive Obsession, Middle Man
Functional	Pure Functions, Side Effects, Hidden Assumptions, Premature Memoization
Testing	Float Assertions, Over-Mocking, Test Without Assertion, Tests Depending on Dates
Error Handling	Crowdstrike Null, Not Sanitized Input, Invalid Id = 9999
Data & Types	Magic Numbers, Mutable Constants, Stringly Typed, Yoda Conditions
Security	Hardcoded Credentials, Insecure Random, Open Redirect
Concurrency	Race Condition, Global State, New Date() in Business Logic
SQL	N+1 Query, SELECT *, God Query, Missing Index
Language-Specific	Mutable Default Args (Python), .unwrap() (Rust), async void (C#), == vs === (JS)

Severity System

	Severity	Action
🔴	Critical	Fix before merge — bugs, crashes, security holes
🟠	Major	Fix this sprint — maintainability and design debt
🟡	Minor	Fix when you're in the area
🔵	Style	Automate with a linter

Installation

This skill uses the Agent Skills open standard — compatible with Claude Code, OpenAI Codex CLI, and any agent that supports SKILL.md.

# Claude Code — install from marketplace
/plugin marketplace add barlevysh/code-smells-skill

# Any agent — clone and copy to your skills directory
git clone https://github.com/barlevysh/code-smells-skill.git
cp -r code-smells ~/.claude/skills/      # Claude Code
cp -r code-smells ~/.codex/skills/       # OpenAI Codex CLI

Usage

Just ask naturally — the agent loads the skill automatically when code review is relevant:

Review this function for code smells.

What's wrong with this Python class?

Find all the testing smells in this file.

Explain what Shotgun Surgery is and show me an example in Go.

Output Format

Each finding is reported as:

### 🔴 Crowdstrike Null
**What it is:** Dereferencing a pointer without checking for null first.
**Why it's harmful:** Causes crashes in production — responsible for the 2024 CrowdStrike outage.
**Where found:** Line 42, processData()

Before:
  int value = data->value; // CRASH if data is NULL

After:
  if (data == NULL) { log_error("null data"); return; }
  int value = data->value;

Skill Structure

code-smells/
├── SKILL.md                          # Master index + workflow
└── references/
    ├── smells-naming.md              # 13 naming & linguistic smells
    ├── smells-structure.md           # 17 structural & complexity smells
    ├── smells-oop.md                 # 29 OOP smells
    ├── smells-functional.md          # 9 functional & pure-code smells
    ├── smells-testing.md             # 15 testing smells
    ├── smells-error-handling.md      # 15 error handling & data smells
    ├── smells-data.md                # Dead code, duplication & paradigm smells
    ├── smells-security.md            # 8 security + 4 concurrency smells
    ├── smells-concurrency.md         # concurrency & state smells
    ├── smells-sql.md                 # 11 SQL smells
    └── smells-language-specific.md   # Per-language gotchas

Uses progressive disclosure — only relevant reference files are loaded per review, keeping context lean.

Why Not Just Use a Linter?

Linters catch formatting and syntax. This skill catches design problems:

A linter can't tell you your class has Feature Envy
A linter won't explain why Shotgun Surgery makes your codebase fragile
A linter won't show you a before/after refactoring in your own language
A linter won't prioritize findings by business impact

Contributing

Found a smell that's missing? PRs welcome.

Fork the repo
Add your smell to the appropriate references/smells-*.md file
Follow the existing format: name, severity, languages, definition, harm, before/after
Update the index in SKILL.md
Open a PR

License

MIT