Python Patterns

python-patterns — Claude Code plugin

A Claude Code plugin that gives the agent on-demand expertise in classical Gang-of-Four and Python-specific design patterns, distilled from faif/python-patterns (42k+ stars). Covers 38 patterns across creational (7), structural (10), behavioral (16), testability (1), fundamental (1), and non-GoF (3) categories. Optimized for three real workflows: writing new code, refactoring existing code, and reviewing pull requests.

Pythonic first. The plugin biases toward plain Python — first-class functions, dataclass, generators, modules-as-singletons, @decorator syntax — and only suggests a classical pattern when the abstraction earns its keep. Singleton, God Object, and deep inheritance are flagged as anti-patterns.

---

Installation

This plugin is published through the cc-garden marketplace. From inside Claude Code:

/plugin marketplace add AABur/cc-garden
/plugin install python-patterns

After installation the python-patterns skill auto-loads on relevant prompts. No slash command — Claude triggers it whenever a Python design or review task matches.

---

How it triggers

Claude consults this plugin when you:

• Name a pattern explicitly: "make this a Factory", "use Strategy here", "refactor with Observer".
• Describe a problem the catalog solves: "I have multiple algorithms for…", "construct an object with 12 fields", "decouple subsystem X", "notify subscribers", "undo/redo", "state machine".
• Ask for design or code review: "is this Pythonic?", "this looks over-engineered", "code review please", "is this the right abstraction?".
• Are about to introduce an anti-pattern: "make this a Singleton" triggers a redirect to Borg or a module-level value, with reasoning.

It stays out of the way for framework-specific tasks (Django, FastAPI, SQLAlchemy idioms), pandas/numpy data work, and one-shot scripts where patterns add nothing.

---

What it does

• New code. Suggests the right pattern with a canonical implementation, plus the Pythonic alternative if one applies.
• Refactoring. Identifies code that warrants a pattern (long if isinstance chains, many constructor variants, hand-rolled event hooks) and proposes a step-by-step refactor recipe.
• Code review. Flags pattern misuse and anti-patterns, distinguishing between legitimate use and gratuitous OOP ceremony.

---

Pattern catalog

Loaded on demand through progressive disclosure — only the relevant pattern's reference file enters context.

Creational (7)

abstract_factory · borg · builder · factory · lazy_evaluation · pool · prototype

Structural (10)

3-tier · adapter · bridge · composite · decorator · facade · flyweight · front_controller · mvc · proxy

Behavioral (16)

chain_of_responsibility · catalog · chaining_method · command · iterator · mediator · memento · observer · publish_subscribe · registry · servant · specification · state · strategy · template · visitor

Testability

dependency_injection

Fundamental

delegation_pattern

Other (non-GoF)

blackboard · graph_search · hsm

Anti-patterns (flagged, not implemented)

Singleton · God Object · inheritance overuse — see skills/python-patterns/references/anti-patterns.md.

---

Layout

python-patterns/
├── .claude-plugin/plugin.json
├── README.md
└── skills/python-patterns/
    ├── SKILL.md                       # router (always loaded body)
    └── references/
        ├── selection-guide.md         # problem → pattern lookup
        ├── pythonic-alternatives.md   # when classical is overkill
        ├── anti-patterns.md           # Singleton / God Object / inheritance
        ├── creational/                # 7 files
        ├── structural/                # 10 files
        ├── behavioral/                # 16 files
        ├── testability/               # dependency_injection
        ├── fundamental/               # delegation_pattern
        └── other/                     # blackboard, graph_search, hsm

---

Attribution

Pattern descriptions, canonical examples, and "Examples in Python ecosystem" notes are adapted from faif/python-patterns (MIT). Anti-pattern guidance derives from the same project's README.

The plugin's editorial bias — Pythonic-first, anti-pattern-aware, refactor-ready — is authored for this distribution.

---

License

MIT. See upstream faif/python-patterns LICENSE for the source material.