Configuration framework adding 30 slash commands, 20 AI agents, and 7 behavioral modes to Claude Code, covering the full dev lifecycle from brainstorm to deploy.
Before you start
Claude Code installed and running. SuperClaude installs into your Claude Code config directory. Know where ~/.claude/ is before proceeding.
SuperClaude installs by copying its config files into your Claude Code settings directory. The installer script handles the copy, it merges into your existing settings rather than overwriting them.
SuperClaude touches your CLAUDE.md and settings.json. Before installing, ask Claude to evaluate what it will change and whether that conflicts with anything you already have configured.
SuperClaude ships with named personas: architect, frontend, backend, security, and others. Each has a focused system prompt and commands it handles well. Invoke with its slash command and give it a task in its domain.
Persona files are plain markdown in `~/.claude/personas/`. Open any one and add stack-specific rules. If you work in Next.js + Supabase, add that context to the frontend persona so it stops suggesting raw SQL when you need Supabase client calls.
Any task you run the same way more than twice is a candidate for a custom command. Add it to `~/.claude/commands/` as a markdown file. It shows up in the command list on the next Claude Code session.
Key SuperClaude slash commands and flags available after installation.
| Command / Skill | What it does | When to use it |
|---|---|---|
/sc:analyze | Deep multi-file codebase analysis with pattern and dependency detection | Understanding an unfamiliar codebase or auditing before a major refactor |
/sc:build | Structured build workflow: plan → implement → verify in sequence | Implementing a feature from spec to working code with a clear audit trail |
/sc:review | Code review covering quality, security, and performance in one pass | Before merging or handing code to another developer |
/sc:design | Architecture design with explicit trade-off analysis and documentation | Planning a new system or major refactor before writing any code |
/sc:debug | Systematic debug loop: reproduce → isolate → fix → verify | A bug that hasn't yielded to normal inspection |
/[persona] | Invoke a named persona (architect, frontend, backend, security, etc.) with domain-focused reasoning | Narrowing Claude's focus to a specific discipline for a task in that domain |
--ultrathinkadvanced | Extended chain-of-thought reasoning before answering | Hard algorithmic or architectural problems with many trade-offs |
--magic | UI/UX generation mode optimised for component and design work | Building interfaces where design quality matters from the first pass |
~/.claude/personas/ | Editable persona markdown files, add stack-specific conventions to any persona | When the default persona keeps suggesting things that don't fit your stack |
~/.claude/commands/ | Custom slash command directory, any markdown file here becomes a command | A workflow you run the same way more than twice |