Claude Code v2.1.90 (released April 1, 2026) shipped a new command called /powerup. It's an interactive tutorial that walks you through 10 core capabilities — from talking to your codebase to dialing model effort. Each lesson is concise and hands-on.
But once you finish a lesson, the natural question is: what now? How do you actually put this into practice on a real project?
I went through all 10 lessons and matched each one with a specific open-source skill or tool from TokRepo that extends what you just learned. Think of this as a companion guide — /powerup teaches the concept, and these tools help you apply it at scale.
1. Talk to your codebase → LLM Wiki Memory Upgrade Prompt
What /powerup teaches: You can reference files with @filename, point to specific lines, and have Claude reason about your codebase structure. It's the foundation of everything else.
What takes it further: The LLM Wiki Memory Upgrade Prompt gives your agent a structured memory schema — a personal wiki that persists across sessions. Instead of re-explaining your architecture every time, the agent already knows your project layout, naming conventions, and key decisions. It turns "talk to your codebase" from a one-shot interaction into a continuous relationship.
2. Steer with modes → Claude Code Templates (600+ Agents, Commands & MCPs)
What /powerup teaches: Shift+Tab toggles between plan mode and auto mode. You can steer how aggressively Claude acts — from careful planning to full autonomy.
What takes it further: Modes are powerful but generic. Claude Code Templates gives you 600+ pre-configured agent setups, slash commands, and MCP integrations. Instead of manually switching modes per task, you get purpose-built workflows — a code reviewer agent that always plans first, a refactoring agent that runs autonomously, and so on.
3. Undo anything → CC Status Board (Smart Status Bar)
What /powerup teaches: /rewind rolls back to any previous state. Double-tap Escape cancels the current operation. You always have an escape hatch.
What takes it further: The problem is knowing when to hit rewind. CC Status Board adds a smart status bar that shows what Claude is doing in real time — which files it's touching, how many tokens it's burning, and whether it's going off track. You catch problems before they need rewinding.
4. Run in the background → Awesome n8n Workflow Automation Templates
What /powerup teaches: You can spin up background tasks with /tasks and let Claude work on multiple things in parallel. It's like having extra hands.
What takes it further: Background tasks inside Claude are great, but some work belongs outside your editor entirely. Awesome n8n is a curated collection of workflow automation templates — CI/CD triggers, Slack notifications on PR reviews, scheduled code quality scans. Pair Claude's background tasks with n8n for anything that should run on a schedule or in response to external events.
5. Teach Claude your rules → Awesome Cursor Rules
What /powerup teaches: CLAUDE.md files and /memory let you define project-specific rules — coding standards, forbidden patterns, architectural constraints. Claude follows them automatically.
What takes it further: Writing good rules from scratch is hard. Awesome Cursor Rules collects 30+ stack-specific rule templates (React, Python, Go, Rust, and more). They were written for Cursor, but the format translates directly to CLAUDE.md. Fork a template, adapt the rules, and you skip the blank-page problem entirely.
6. Extend with tools → Chrome MCP Background Proxy
What /powerup teaches: MCP (Model Context Protocol) servers let Claude use external tools — databases, APIs, browsers. /mcp manages your connections.
What takes it further: The single biggest MCP pain point is browser automation stealing window focus. Every time Claude uses a Chrome MCP tool, the browser pops to the foreground and interrupts your work. Chrome MCP Background Proxy solves this by routing MCP calls through a background proxy. Claude browses, scrapes, and tests without ever stealing your screen.
7. Automate your workflow → Lark CLI Skill: Skill Maker
What /powerup teaches: Skills are reusable capabilities you can teach Claude — custom slash commands, hooks that fire on events, repeatable procedures stored as markdown.
What takes it further: Skill Maker is a meta-skill: it helps you create new skills faster. Describe what you want in plain English, and it generates the skill file with proper frontmatter, parameter definitions, and step-by-step instructions. Instead of hand-writing skill markdown, you scaffold it in seconds.
8. Multiply yourself → GPT Researcher (Autonomous Research Agent)
What /powerup teaches: Subagents let Claude delegate work to child agents. /agents manages them. You can parallelize research, code review, and testing across multiple agents.
What takes it further: Subagents are only as good as their specialization. GPT Researcher is a purpose-built autonomous research agent that searches the web, aggregates sources, and produces structured reports. Use it as a subagent when Claude needs to answer questions that go beyond your codebase — competitor analysis, library comparisons, or technical deep dives.
9. Code from anywhere → Tabby (Self-Hosted AI Coding Assistant)
What /powerup teaches: /remote-control and /teleport let you use Claude Code from your phone, another machine, or any browser. Your dev environment goes wherever you go.
What takes it further: Remote Claude is great, but sometimes you want local AI completions too — especially when you're offline or working on sensitive code. Tabby is a self-hosted AI coding assistant that runs on your own hardware. Use Claude remotely for heavy tasks, Tabby locally for fast inline completions. They complement each other.
10. Dial the model → Agent Skills Standard (Cross-Platform AI Skills)
What /powerup teaches: /model switches between Claude models. /effort adjusts how hard Claude thinks. You can optimize for speed or depth depending on the task.
What takes it further: Once you start switching models, you realize your skills and prompts might be model-dependent. Agent Skills Standard defines a cross-platform skill format that works across Claude, Codex, Gemini, and others. Write your skills once using this standard, and they work regardless of which model you've dialed in — no rewrites needed when you switch.
Keep going
/powerup is a great starting point, but the real leverage comes from the tools and skills the community builds on top. All the resources linked above — and hundreds more — are browsable at tokrepo.com. Filter by your stack, grab what fits, and skip the setup work that someone else already solved.
If you've built a skill or workflow that pairs well with a /powerup lesson, drop a link in the comments. The best dev tools are the ones people actually share.
Top comments (0)