Development · Framework
The Shorthand Guide to Everything Claude Code
A practical reference guide and configuration framework for Claude Code power users.
A detailed configuration framework and workflow guide for Claude Code based on ten months of intensive daily use. It covers the setup and architecture for skills, lifecycle hooks, modular subagents, custom rules, and Model Context Protocol (MCP) servers. It emphasizes strict context-window management, LSP integration, and CLI developer ergonomics.
The framework
Skills: shorthand prompts stored in ~/.claude/skills (broad workflows) vs commands in ~/.claude/commands (quick executable prompts). Can be chained in one prompt (e.g. /refactor-clean, then /test-coverage, then /e2e). Example: a codemap-updater skill to refresh code navigation maps without burning context.
Hooks: trigger-based automations on tool calls/lifecycle events — PreToolUse, PostToolUse, UserPromptSubmit, Stop, PreCompact, Notification. Example: a PreToolUse hook reminding to use tmux for long-running bash commands. Use the /hookify plugin to create hooks conversationally instead of writing JSON.
Subagents: delegate tasks with limited scopes to background/foreground processes defined in ~/.claude/agents (planner, architect, tdd-guide, code-reviewer, security-reviewer, build-error-resolver, e2e-runner, refactor-cleaner). Scope tools/MCPs per subagent.
Rules: either one CLAUDE.md or a modular ~/.claude/rules folder (security.md, coding-style.md, testing.md, git-workflow.md, agents.md, performance.md). Example rules: no emojis in codebase, no purple hues in frontend, test before deploy, modular over mega-files, never commit console.logs.
MCPs: connect Claude to external services (Supabase MCP runs SQL upstream; Chrome built-in MCP drives your browser). CRITICAL context-window advice: disable all unused MCPs via /plugins or /mcp — 200k context can shrink to ~70k with too many tools. Rule of thumb: 20–30 MCPs in config, under 10 enabled, under 80 tools active.
Plugins: bundle skills+MCPs+hooks. Install via 'claude plugin marketplace add <url>' then /plugins. LSP plugins (typescript-lsp, pyright-lsp) give Claude type checking and go-to-definition outside an IDE.
Tips: Ctrl+U deletes whole line; ! bash prefix; @ file search; Shift+Enter multiline; Tab toggles thinking; Esc Esc interrupts/restores code. /fork for parallel non-overlapping tasks; git worktrees for parallel Claude instances without conflicts; tmux (tmux new -s dev / tmux attach -t dev) to watch Claude's long-running commands; mgrep beats ripgrep (local + web search); /rewind, /statusline, /checkpoints, /compact. Set up Claude GitHub Actions PR review. Use sandbox mode for risky operations (--dangerously-skip-permissions to disable).
Editors: prefers Zed (Rust, fast, Agent Panel tracks Claude's file changes, CMD+Shift+R command palette, vim mode) — split screen with terminal, Ctrl+G to open the file Claude is editing, enable autosave. VSCode/Cursor also viable via \ide or the extension.
His plugin stack (keep 4–5 enabled): ralph-wiggum, frontend-design, commit-commands, security-guidance, pr-review-toolkit, code-review, feature-dev, explanatory-output-style (claude-code-plugins marketplace); typescript-lsp, pyright-lsp, hookify, code-simplifier, context7 (claude-plugins-official); mgrep (Mixedbread-Grep). MCP servers configured: github, firecrawl, supabase, memory, sequential-thinking, vercel, railway.
What it's for
- Configure custom Claude Code skills and lifecycle hooks for automated developer workflows
- Set up specialized subagents with scoped tool access for testing, architecture, and security reviews
- Manage MCP server overhead to preserve Claude's context window
- Pair Claude Code with terminal multiplexers and fast editors like Zed for parallel development
Key points
- Skills vs. Commands setup guide for reusable and chainable developer prompts
- Event-driven lifecycle hooks (PreToolUse, PostToolUse, UserPromptSubmit) with conversational hook generation
- Scoped subagent architecture for planner, architect, TDD, and reviewer roles
- Modular rules folder structure replacing monolithic CLAUDE.md files
- Context preservation rules of thumb limiting active MCP tools to under 80
- Recommended plugin stack including language LSPs, mgrep, and review toolkits
- CLI shortcuts, tmux workflows, and Git worktree strategies for running parallel instances
Pricing
Free
Details
- Type
- Framework
- Category
- Development
- Topic
- AI Coding Workflows
- Open source
- No
- Platforms
- Command line
- For
- Developers and engineers using the Claude Code CLI interface.
- Tags
- claude code, anthropic, mcp, developer tools, cli, prompt engineering, workflow
Why it was saved
- @affaan: A concrete, opinionated reference for structuring skills, hooks, subagents, rules, MCPs, and plugins — including the critical context-window management advice.
Related
Practical eight-step setup and workflow guide for building MVPs with Cursor
Battle-tested Claude Code workflow tips directly from the tool's creator
An 8-step framework for building apps with AI using structured documentation
Context engineering framework for making AI coding agents reliable at scale
Curated collection of 160+ specialized modular subagents for Claude Code
Agent-first control plane backed by 2,400+ curated AI developer skills
Updated