simpleclaude

SimpleClaude

A practical minimalist AI assistant framework that transforms complex AI interactions into natural conversations. Comes with a set of user-commands, sub-agents, hooks, and utilities designed for real-world software development tasks. Completely dogfooded by its creator for building production software in the workplace and for personal projects. Expect frequent updates, enhancements, and breaking changes.

Installation

Option 1: Direct from GitHub (Recommended)

Install SimpleClaude plugins directly from the commandline:

claude plugin marketplace add https://github.com/kylesnowschwartz/SimpleClaude
claude plugin install sc-core
claude plugin install sc-hooks
claude plugin install sc-output-styles
claude plugin install sc-extras
claude plugin install sc-research
claude plugin install sc-skills
claude plugin install sc-docs
claude plugin install sc-mods

What's included:

• sc-core: Intent-based commands and specialized agents for planning, implementation, exploration, and review
• sc-hooks (optional): Auto-formatting and lint checks on stop, plus tool monitoring (steers GitHub WebFetch to the gh CLI)
• sc-output-styles (optional): Curated output styles — personality-driven (Linus, Austen, Lovelace, Ousterhout, Starfleet, Mayo Clinic, Ogg) and tail-first replies for skimmers (Bottom Line, plus a Simplified Technical English variant)
• sc-extras (optional): The sc-review-and-refactor skill, which routes review and refactoring work to a named technique (safe refactoring, test-quality review, verifying findings and claims, root-cause analysis, PR review threads); general diff review goes to the built-in /code-review. Also adversarial analysis, context wizards, and clean-room design
• sc-research (optional): Research agents for library source (via opensrc), GitHub search (via gh), and web research
• sc-skills (optional): Hypothesis testing, Socratic thinking, functional thinking, file querying, image generation, and coding rules
• sc-docs (optional): Writing and diagram skills: human-writing for plain prose in Slack messages, emails, PRs, issues, tickets, and docs; cold-reader, which checks that a persisted document makes sense to a reader with no project context; and sc-mermaid, which picks the mermaid diagram type that fits (sequence, state, ER, gantt, and more) and validates it with mmdc. Both writing skills lint drafts with Vale (brew install vale)
• sc-mods (optional): Claude Code mods. Draws mermaid diagrams in replies as Unicode text in the terminal, adds prompt-jump buttons above the prompt, and a copy button under each reply. See plugins/sc-mods/README.md

Updating

Via Claude Code CLI

claude plugin marketplace update simpleclaude
claude plugin install sc-core@simpleclaude
claude plugin install sc-hooks@simpleclaude
claude plugin install sc-output-styles@simpleclaude
claude plugin install sc-extras@simpleclaude
claude plugin install sc-research@simpleclaude
claude plugin install sc-skills@simpleclaude
claude plugin install sc-docs@simpleclaude
claude plugin install sc-mods@simpleclaude

Quick Start

# Core development commands
/sc-core:sc-plan "How should I add authentication to this app?"
/sc-core:sc-work "Add JWT authentication with login/logout"
/sc-core:sc-explore "How does the current database layer work?"
/sc-core:sc-review "Check security vulnerabilities in auth module"
/sc-core:sc-workflow "Start structured development for OAuth support"

# Review and refactoring
/code-review high my-branch               # Built-in general review: level first, then target
/sc-extras:sc-review-and-refactor did this branch do everything the ticket asked?   # Requirements trace
/sc-extras:sc-review-and-refactor resolve the review threads on PR 42              # PR review threads
/sc-extras:sc-review-and-refactor why does the login test fail?                     # Root-cause analysis

# Utilities
/sc-docs:sc-mermaid auth flow             # Pick a diagram type, draw it, validate it

Architecture

SimpleClaude uses a lightweight agent architecture with intent-based commands:

• Command Structure: Intent-based interfaces that understand user goals (plan, work, explore, review, workflow)
• Specialized Agents: Context-specific assistants spawned via Task() calls
• Token-Efficient Design: Agents work in isolation with focused context
• Clean Separation: Commands understand intent, agents execute specialized tasks

Runtime configuration (environment variables) for sc-hooks is documented at plugins/sc-hooks/README.md.

Agent System

Each command automatically spawns appropriate agents:

• sc-code-explorer - Traces execution paths and maps architecture
• sc-code-architect - Designs feature architectures from existing patterns
• sc-code-reviewer - Reviews code for bugs, security, and convention adherence

The optional sc-research plugin adds research agents that Claude picks by description:

• sc-research-github - Searches across GitHub to compare libraries, evaluate projects, and find issues/PRs
• sc-research-repo - Fetches a library's source with opensrc and searches it for docs, APIs, and source code
• sc-research-web - Searches the web for news, blogs, tutorials, and community discussions

Project Structure

SimpleClaude/
├── plugins/
│   ├── sc-core/        # Core plugin: commands & agents
│   ├── sc-hooks/                 # Hooks plugin: auto-format, lint, tool monitoring
│   ├── sc-output-styles/         # Output styles plugin: personality and tail-first reply styles
│   ├── sc-extras/                # Extras plugin: review-and-refactor skill, utility commands
│   ├── sc-research/              # Research plugin: library source, GitHub, and web agents
│   ├── sc-skills/                # Skills plugin: patterns, design, querying
│   ├── sc-docs/                  # Docs plugin: writing and mermaid diagram skills
│   └── sc-mods/                  # Mods plugin: mermaid diagrams, prompt jumps, reply copy
├── scripts/
│   ├── install.rb                # Installation/update script
│   └── update-merman.sh          # Pins the merman release sc-mods downloads
├── docs/                         # Documentation
└── README.md                     # This file

Acknowledgments

SimpleClaude was inspired by SuperClaude which demonstrated effective patterns for natural language command interfaces. SimpleClaude builds on concepts pioneered by the SuperClaude project and I thank NomenAK for putting in the hard yards with SuperClaude - an awesome project with powerful features.