English | 中文 | Codex Branch | Changelog

Awesome Claude Code Configuration

Production-ready configuration for Claude Code. One-command install of global instructions, multi-language coding rules (Python / TypeScript / Go), 21 curated plugins across 10 marketplaces, seven bundled skills, a gradient status bar, and a self-improvement loop that remembers corrections across sessions.

Showcase

• paper-reading skill — Attention Is All You Need
• adversarial-review skill — worked example

Quick Start

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/Mizoreww/awesome-claude-code-config/main/install.sh)

Windows (PowerShell):

irm https://raw.githubusercontent.com/Mizoreww/awesome-claude-code-config/main/install.ps1 | iex

Launches a two-level interactive selector. Append --all / -All to skip the menu and install everything non-interactively. Other flags: --dry-run, --uninstall, --version (PowerShell: -DryRun, -Uninstall, -Version).

  > [5/5] Core                   Global instructions, settings, rules...
    [0/3] Language Rules          Python / TypeScript / Go
    [2/3] Review                  code-review + adversarial-review
    [5/6] Workflow                karpathy, mattpocock-skills, neat-freak...
    [2/2] Integrations            context7, playwright
    [4/5] Design & Content        document-skills, frontend-design, humanizer...
    [0/2] Slides                  frontend-slides, ppt-master
    [0/2] Memory & Lifestyle      claude-mem, claude-health
    [0/1] Storage                 storage-analyzer
    [1/11] Academic Research      paper-reading, deepxiv-cli, researchstudio...
    [0/1] MCP Servers             Lark/Feishu

• Main menu: ↑↓ navigate groups, Enter or → open a group’s sub-menu, q quit. Arrow to Submit and press Enter to install.
• Sub-menu: ↑↓ navigate items, Space toggle, ← or Esc back to main menu (same as pressing Enter on [ Back ]).
• Shortcuts (any level): a all on, n all off, d defaults; in sub-menus these only affect that group.
• The Review group’s adversarial-review and codex are mutually exclusive — selecting one deselects the other.

Core (5) — foundational files, all on by default.

Language Rules (3) — off by default, enable only what your projects use.

Review (3) — adversarial-review and codex are mutually exclusive.

Workflow (6) — planning, iteration, code quality, meta-config.

Integrations (2) — external tools and services.

Design & Content (5) — documents, UI, creative artifacts, text humanization.

Slides (2) — AI slide / PPTX generation, both off by default.

Memory & Lifestyle (2) — session memory and personal productivity, all off by default.

Storage (1) — disk usage analysis, off by default.

Academic Research (11) — training / inference plugins + paper-reading, DeepXiv & ResearchStudio skills, off by default except paper-reading.

MCP Servers (1) — non-plugin MCP integrations, off by default.

Directory Structure

.
├── CLAUDE.md              # Global instructions
├── settings.json          # Permissions, plugins, hooks, model
├── lessons.md             # Self-correction log template (auto-loaded via hook)
├── rules/                 # Coding standards (common + python/typescript/golang)
├── hooks/                 # Statusline with gradient progress bars
├── mcp/                   # MCP server config (Lark-MCP)
├── plugins/               # Plugin catalogue & install guide
├── skills/                # Bundled custom skills
├── docs/                  # Paper summaries, showcases
└── install.sh / install.ps1

Key Mechanisms

• Layered rules — rules/common/ (universal) extended by per-language directories. Each file references a deeper skill for patterns, testing, security.
• Statusline — model, directory, venv, git branch, context window (gradient bar), 5-hour usage countdown. Script at hooks/statusline.sh.
• Self-improvement loop — corrections route to ~/.claude/lessons.md (cross-project) or project MEMORY.md (local). SessionStart hooks re-inject them on startup and after context compaction.
• Plugin catalogue & marketplace URLs — full list with install commands: plugins/README.md.

Settings Defaults

settings.json ships with high-performance defaults. Unknown keys are ignored by older Claude Code; only auto mode is version-gated (installer auto-downgrades to bypassPermissions below 2.1.80).

Smart merge on re-install preserves your overrides for env, permissions.allow, enabledPlugins, hooks.SessionStart, and statusLine. Third-party plugins in your enabledPlugins that are outside this catalogue are left untouched.

Customization

• Add a language: create rules// extending rules/common/
• Add a skill: place in skills//SKILL.md
• Adapt CLAUDE.md: tune for your shell, package manager, project context

Acknowledgements

• Claude Code in Action — Anthropic Academy’s official course
• Working for 10 Claude Codes by Hu Yuanming — multi-instance patterns
• Harness Engineering by OpenAI
• Anthropic Engineering / OpenAI Engineering
• Claude Code Best Practice by shanraisshan
• Claude How To by luongnv89

License

MIT