CLI Maxxing

Everything you need to start working with AI-powered development tools; one command per step.

Note

Heads up — this was fidgetcoded. I'm no expert, but I tried — nominally — my best. Getting this stack running end-to-end on a clean machine took a lot of trial and error, and it still might not be perfect.

---

fidg·et·cod·ing \ ˈfij-ət-ˌkō-diŋ \ verb · fidgetcoded; fidgetcode
1 : to code as self-regulation — the ADHD dopamine loop
2 : to build software through relentless tinkering by someone with no formal training and too many browser tabs open

---

Read everything carefully as you go (or don't, entirely your call). If something breaks mid-install, don't panic (or do, also your call):

1. Get Claude CLI up first (Step 1). Once it's running, you can paste any error straight into it and Claude will help you troubleshoot EVERYTHING downstream.
2. Re-run the install step. A lot of first-pass failures quietly clear themselves on the second try — re-running is almost always the right first move before anything fancier.

The Trilogy

This is one of three repos in the cli-maxxing stack:

Repo	What it does
cli-maxxing	This repo — Foundation — Claude Code, shell aliases, FidgetFlo, dev tools, productivity MCPs
creativity-maxxing	Design skills + video/audio pipeline
task-maxxing	Two-way task sync — Obsidian ↔ Morgen (Notion dropped 2026-05-04) (requires 2ndBrain-mogging)

Install cli-maxxing first. creativity-maxxing and task-maxxing can be installed in either order after that.

Quick Navigation

	Section	What it does	Time
Before You Start	Requirements	What you need before running anything
How It Works	Overview	How the steps fit together
Install Everything	One-shot	Run all steps at once	~20 min
Keyboard + Command Cheat Sheet	Commands & Shortcuts	Hotkeys, typing, and commands for your terminal
Step 1	CLI Tools	Git, Node.js, Claude Code, shell aliases — the foundation	~10 min
Step 2	Bonus Software	Ghostty (terminal) + Arc (browser) — optional but highly recommended	~4 min
Step 3	Developer & Utility Tools	Adds file converters, search, utilities, and no-flicker mode	~3 min
Step 4	FidgetFlo	Multi-agent orchestration — swarms, hives, persistent memory, Opus-locked	~3 min
Step 5	Productivity Tools	Notion + Granola + n8n + GCal + Morgen + Motion + Playwright + SwiftKit + Superhuman + Google Drive + Vercel (pick what you use; Morgen recommended)	~5 min
Step 6	Telegram	Message Claude from your phone via Telegram bot	~2 min
Step 7	GitHub	GitHub MCP + /gitfix + /recon skills — repos, issues, PRs, code search, full-repo doc sync, pre-build prior-art recon (MCP requires PAT)	~2 min
Step 8	Safety Check	Security auditing — scan any project for vulnerabilities + full MCP security checks	~2 min
Final Step	Status Line	Final config — status indicators for active swarms, vault, MCP	~2 min
You're Ready	Start here after setup	Your daily command and what to do next
Staying Up to Date	Update command	Re-run everything, catch new steps
Uninstall	Remove everything	Reverses all steps, cleans up tools and config

---

Before You Start

Back to top

Important

You need a paid claude.ai subscription before anything below is useful.

The Claude Code CLI installs for free — but USING Claude requires a paid plan. If you don't have one yet, sign up first (Claude Pro is the minimum tier at $20/mo; Max is recommended for heavy use). Otherwise the installer downloads ~500 MB of tooling that has nothing to talk to.

The installer will prompt you upfront — press Enter if you're subscribed, Ctrl+C to bail and sign up first.

• Your computer needs to be from roughly 2020 or later (macOS Big Sur+ or a recent Linux).
• You need an internet connection since the scripts download everything live.
• Don't run it as root. Just open your terminal normally and paste the command.
• If anything is already installed on your machine, the scripts will detect that and skip it automatically.

Windows: Not supported. macOS and Linux only (frankly, I'm not 100% certain Linux is supported).

---

How It Works

Back to top

Run the steps in order. Each one builds on the last.

Step 1 — CLI Tools is the only part that feels "techy." This step gets the bare essentials on your machine so Claude (your AI assistant) can run — Git, Node.js, Claude Code, and shell aliases. You paste one command and it handles the rest, but there are a few manual steps after it finishes, like logging into Claude. This is the most hands-on part of the entire process. After Step 1, you can ask Claude questions at any point. If something doesn't make sense, just ask. That's the whole point.

Step 2 — Bonus Software is optional but highly recommended. It installs two tools that make your day-to-day flow noticeably smoother: Ghostty, a GPU-accelerated terminal with Cmd+Click support for URLs and file paths, and Arc, a Chromium-based browser with a sidebar instead of a tab bar, Spaces for context switching, and built-in ad blocking. You don't need either to continue — everything works in any terminal and any browser — but if you want the best terminal + browser setup, this is it.

Step 3 — Developer & Utility Tools is where you install the rest of your development tools. Things like file converters, search tools, and utilities. You run this from your terminal after Step 1 is done. Much more straightforward.

Step 4 — FidgetFlo is where you set up FidgetFlo, the multi-agent orchestration layer that turns Claude into a full team of AI agents — /fswarm, /fmini, /fhive, persistent memory, Opus-locked.

Step 5 — Productivity Tools connects Claude to your productivity tools — notes, calendars, email, meetings, workflows, browser automation, deployments, and hosted toolkits. Pick the ones you use: Notion, Granola, your own n8n instance, Google Calendar, Morgen (recommended), Motion Calendar, Playwright, SwiftKit, Superhuman, Google Drive, or Vercel. All optional, install only what you need.

Step 6 — Telegram connects Claude to Telegram so you can message it straight from your phone. You create a free bot through Telegram (takes about two minutes), the script handles the rest, and then you use ctg or cbraintg to launch Claude with Telegram connected — messages show up in your session in real time. This step is completely optional; everything else works without it.

Step 7 — GitHub is the GitHub bundle — for developers. It installs the GitHub MCP so Claude can read and write your repos, issues, pull requests, and search code across your GitHub organizations (requires a GitHub Personal Access Token), plus the /gitfix skill that reads every file in a repo and fixes any drift between your code and your docs, and the /recon skill that sweeps GitHub + the web for prior art before you build something new. Skip this step if you don't use GitHub with Claude.

Step 8 — Safety Check installs a security auditing skill that lets Claude scan any project for exposed keys, missing rate limiting, input sanitization gaps, dependency vulnerabilities, and more. Just point Claude at a project and ask it to run a safety check. It catches the stuff that slips through code review.

Final Step — Status Line is the wrap-up. It installs a custom status line that shows you what's active at a glance — your vault, MCP connection, design tools, and any running swarms, mini swarms, or hive-minds.

After the Final Step, head to You're Ready — it tells you the one command you need going forward and what to do next.

Between Steps 1 and 2, make sure to read the Keyboard + Command Cheat Sheet so you know how to type, navigate, and use hotkeys in your terminal.

Already done with everything? Use the Staying Up to Date command to catch any new steps or updates that have been added since your last visit.

If you ever want to start fresh or remove everything this setup installed, there's a one-command Uninstall that reverses all steps. It won't touch your Obsidian vault or notes.

Already have Claude Code installed?

If you already have Claude Code working on your machine, you can skip Step 1 entirely. Just jump straight to Step 3. Everything will work the same. You can paste the install commands directly in your terminal, or if you prefer, download this repo as a ZIP from GitHub, unzip it, and tell Claude to run the scripts from whatever folder they landed in.

Bonus

Tip

Want to get better at using the terminal in general? Check out Terminal Academy, a gamified way to learn terminal commands and workflows. It makes the learning curve way less painful.

---

Install Everything

Back to top

If you already know your way around a terminal and just want everything installed at once:

Important

Paste this into your terminal:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/install.sh)

This runs Steps 1, 2, 3, 4, 8, and the Final Step automatically, plus both bonuses (Ghostty and Arc Browser). Arc is macOS-only and will be skipped on Linux. Everything is idempotent — already-installed tools are skipped.

---

Keyboard + Command Cheat Sheet

Back to top

This is a quick reference for terminal hotkeys, typing basics, launching Claude, and useful commands. Read this before starting the steps, especially if you're new to working in a terminal.

Open the full Cheat Sheet →

Here are the commands you'll use most:

Command	What it does
cskip	Start with all permissions skipped (fastest, no prompts)
cbrain	Jump straight into your 2ndBrain vault with permissions skipped (requires 2ndBrain-mogging)
Shift+Tab	Toggle permissions on/off mid-session without restarting
/fswarm *write task here*	Launch a 15-agent FidgetFlo swarm — just describe what you want in plain English after /fswarm
/fmini *write task here*	Launch a compact 5-agent FidgetFlo swarm — same power, tighter team. Describe your task after /fmini
/w4w	Maximum attention to detail mode — word for word, line for line. No skipping, no summarizing, zero regard for credit burn
/concise	Strip default-LLM fluff — no headers on simple Qs, no "great question", no scaffolding. Auto-suspends for copy/scripts/decks
/bullets	Crush a paragraph into the shortest scannable bullets — core facts only, zero fluff. Always bulletizes. For handoffs nobody should read as a blob. Also works in plain English ("bullet this")
Ctrl+C	Stop whatever is running or exit Claude
/resume	Pick up right where you left off — reloads your last session's context

Everything else — aliases, slash commands, natural-language tools, troubleshooting — is in the full Cheat Sheet.

---

Step 1 - CLI Tools

Back to top

This step installs the minimum needed to get Claude Code running on your machine. It's the most hands-on part of the setup — everything after this can be handled by Claude itself.

macOS / Linux

Open Terminal: Cmd+Space → "Terminal" on Mac, or Ctrl+Alt+T on Linux.

Important

Paste this in and hit Enter:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-1/step-1-install.sh)

The script will ask for your Mac password to install system tools. When you see Password:, type it and hit Enter — you won't see the characters, that's normal.

Xcode Command Line Tools dialog (Mac). Early in the install, macOS pops up a system dialog to download Apple's Command Line Tools. Click Install (and Agree on the license prompt) the moment it appears. The dialog will claim "About an hour remaining" — ignore it. The actual download is ~3–5 minutes and the timer never refreshes. Just leave the dialog open, let it finish, and the install script picks right back up.

What This Step Installs

Tool	What it does
Xcode CLT (Mac) / build-essential (Linux)	Build tools that other installers need.
Homebrew (Mac) / apt or dnf (Linux)	Package manager — installs other software.
Git	Tracks and manages code changes.
nvm + Node.js (v18+)	Node Version Manager + the JavaScript runtime Claude Code needs.
Claude Code	Your AI coding assistant. The main tool.
Shell aliases	cskip, cc, ccr, ccc — faster ways to launch Claude.
ctg	Launches Claude with Telegram connected from any directory.
cbrain	Launches Claude pointed at your Obsidian vault. Installed by 2ndBrain-mogging — available after you set up the vault.
cbraintg	cbrain + Telegram. Also installed by 2ndBrain-mogging.

After the script finishes

1. Close this terminal and open a new one. Homebrew, nvm, and the new aliases only load in a fresh shell. claude won't be on your PATH until you restart.
2. Run claude --version — you should see something like 2.1.112 (Claude Code). If you get "command not found," try source ~/.zshrc (or ~/.bashrc). Still stuck? See Troubleshooting.
3. Press Ctrl+C to exit Claude, then run cskip — this starts Claude in auto-approve mode (no permission prompts), the recommended way to run the rest of the setup.

Shift+Tab toggles between permission-asking mode and auto-approve mode inside any running Claude session — no need to restart.

Set Up Your Claude Account

Sign up at claude.ai before Step 3. Claude Code is free to install but requires a paid plan to actually use.

Why Claude? Anthropic consistently ships the strongest model for complex reasoning + code, and they lead on safety research instead of chasing hype. Claude is careful, honest, and doesn't BS you when it doesn't know something — it's the smartest tool in the room, built by people who care about getting this right.

Subscription Plans

Plan	Cost	What you get
Claude Pro	$20/month	Everyday tasks — writing code, editing files, Q&A. You'll hit limits on long sessions.
Claude Max 5x	$100/month	5× Pro usage. Best for daily users and multi-step workflows.
Claude Max 20x	$200/month	20× Pro usage. Full codebase refactors, multi-agent swarms, all-day sessions.

My recommendation: Start with Pro ($20/month). If you hit rate limits, upgrade to Max — you'll know pretty quickly which tier fits your workflow.

That's it for Step 1. Continue to Step 2 — Bonus Software for Ghostty + Arc, or jump straight to Step 3 — Developer & Utility Tools.

---

Step 2 - Bonus Software

Back to top

Optional but highly recommended. Installs Ghostty (GPU-accelerated terminal with clickable links + g2/g4 window tiling) and Arc (sidebar-tab browser with Spaces, split view, built-in ad blocking). Everything else in this setup works without them — skip if you're happy with your current terminal + browser.

Run both at once

Important

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-2/step-2-install.sh)

Prefer to do one at a time? Run either script individually:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-2/ghostty-install.sh)
bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-2/arc-install.sh)

---

Ghostty Terminal

GPU-accelerated (Metal on Mac, OpenGL on Linux). Cmd+Click any URL or file path to open it. Comes tuned: JetBrains Mono font, dark theme, tabbed windows, plus g2/g4 aliases to tile 2 or 4 Ghostty windows across your screen in one keystroke — great for running multiple Claude sessions side-by-side.

Docs: ghostty.org/docs · Config: ~/Library/Application Support/com.mitchellh.ghostty/config (Mac) or ~/.config/ghostty/config (Linux)

Script failed? Install Ghostty manually

1. Go to ghostty.org/download → Download for macOS (or your Linux package).
2. Open the .dmg, drag Ghostty.app into Applications, eject the image.
3. Open Ghostty (Cmd+Space → Ghostty) → click Open on the macOS warning.
4. Re-run the Ghostty install script above to pick up the font + theme + g2/g4 config. It's idempotent.

Important

One more step — grant Ghostty Full Disk Access (macOS). Without this, Ghostty silently fails to read most user files. The install script prints these instructions at the end, but here's the heads-up:

1. Open System Settings (Cmd+Space → "System Settings")
2. Go to Privacy & Security → Full Disk Access
3. Toggle Ghostty ON. If it's not in the list: click +, navigate to /Applications, select Ghostty.app, then enable the toggle.
4. Quit Ghostty fully (Cmd+Q) and reopen it.

Shortcut: re-run the install script with --open-fda to jump straight to the right pane in System Settings.

---

Arc Browser

Chromium-based with a sidebar instead of a tab bar, Spaces for context switching, split view, built-in ad blocking, and a Cmd+T command bar that replaces the URL bar. Imports from Chrome in 30 seconds — bookmarks, passwords, history, and all your extensions come along.

Download: arc.net · macOS + Windows only (Linux skips this step automatically)

Script failed? Install Arc manually

1. Go to arc.net → Download Arc (auto-detects your OS).
2. Open the downloaded file → drag Arc.app into Applications (Mac), or run the .exe (Windows).
3. Open Arc → click Open on the macOS warning.
4. Sign up for a free Arc account (required for sync) and accept the Chrome import prompt.
5. Set Arc as default browser when asked — or later via System Settings → Desktop & Dock → Default web browser.

---

Step 3 - Developer & Utility Tools

Back to top

Installs the dev tools Claude leans on when working on your projects — file converters, search tools, PDF/doc utilities — plus no-flicker mode (fullscreen rendering, no screen jumping) and a memory auto-save hook (Claude remembers context between sessions).

Run Step 3

In a fresh terminal, launch Claude in auto-approve mode:

cskip

First time? Claude opens a browser to log in with your Anthropic account. Once you're inside the session, paste this:

Important

Paste this into your Claude session:

run this command to install my dev tools: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-3/step-3-install.sh)

Claude runs the install. If it asks you to restart your terminal, close the window, reopen, cskip again, and tell Claude to pick up where it left off.

Why cskip? Auto-approve mode lets Claude run this install (10+ tools) in one smooth pass instead of asking permission for every action. Press Shift+Tab anytime to toggle it off without restarting. Normal claude mode is fine for daily use — cskip is just for setup.

What This Step Installs

Tool	What it does
Python 3 + pip	Runs Python scripts and tools.
Pandoc	Converts Word / PowerPoint / Markdown between formats.
xlsx2csv	Converts spreadsheets to readable CSV.
pdftotext	Extracts text from PDFs.
jq	Reads and edits JSON config files.
ripgrep	Fast code search — Claude Code uses it internally.
tree	Shows your folder structure visually.
fzf	Fuzzy-finder for files and commands.
wget	Downloads files from the web.
weasyprint	Converts HTML to PDF (Claude uses this to generate docs).
No-flicker mode	Fullscreen rendering in Claude Code — screen stops jumping while Claude works. Scroll speed set to 3.
Memory auto-save hook	Claude auto-saves important context when a session ends, so it remembers across sessions.

No-flicker mode details

Research-preview feature (Claude Code v2.1.89+). Without it, the input box bounces while Claude streams output. With it on, input stays pinned to the bottom and everything updates cleanly.

Sets CLAUDE_CODE_NO_FLICKER=1 and CLAUDE_CODE_SCROLL_SPEED=3 in your shell profile. Scroll with mouse wheel, PgUp/PgDn, or Ctrl+Home/End. Click collapsed tool results to expand, click URLs to open.

To disable: remove both export lines from ~/.zshrc (or ~/.bashrc) and restart your terminal. Or set CLAUDE_CODE_NO_FLICKER=0 for a temporary off-switch.

Memory auto-save hook details

A "stop hook" fires every time you end a Claude session (Ctrl+C or /exit). Claude reviews the conversation and saves anything important to memory — decisions, preferences, project context. Next session, it already knows. Nothing to configure; just keep using Claude and memory builds up over time.

---

Step 4 - FidgetFlo

Back to top

FidgetFlo 💚 is a fork of ruvnet's Ruflo, tuned for Claude Opus 4.7. It turns Claude Code from a single assistant into a coordinated team of AI agents: multi-agent swarms on demand, persistent memory, self-healing workflows, and all agents Opus-locked by default (no silent downgrade to Haiku/Sonnet). (💚 = built by fidgetcoding.)

Run Step 4

Still in a cskip session? Good. Paste this:

Important

Paste this into your Claude session:

run this command to set up FidgetFlo: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-4/step-4-install.sh)

If Claude tells you to restart your terminal, close the window, reopen, cskip again, and tell Claude to pick up where it left off.

Commands

Command	What it does
/fswarm <task>	Launches 15 agents in parallel — architect, backend devs, testers, security auditor, etc. Brute-force execution.
/fmini <task>	Launches 5 agents — architect, dev, tester, reviewer, researcher. Tighter team for focused work.
/fhive <goal>	Queen agent takes full control — decides what workers to spawn and how to coordinate. Set the goal, step back.
/w4w	Word-for-word, line-for-line. Maximum attention, zero skipping.
/concise	Default chat shape — strips fluff, scaffolding, and sycophancy. Suspends for copywriting deliverables.
/bullets	Crush a prose blob into the shortest bullets — one fact each, zero fluff, signal kept. Always bulletizes; for handoffs.

Thinking tiers

Subagents inherit the parent's model (Opus stays Opus) but not its /effort setting — the slider doesn't tether to spawned agents. These tier variants bake the trigger into each Agent prompt:

Tier	Mini (5 agents)	Swarm (15 agents)	Trigger appended	Budget
0	/fmini	/fswarm	(none)	0
1	/fmini1	/fswarm1	Think.	~4k
2	/fmini2	/fswarm2	Think hard.	~10k
3	/fmini3	/fswarm3	Think harder.	~31k
max	/fminimax	/fswarmmax	Ultrathink.	~32k

Natural-language aliases work too: "hard"/"deep" → tier 2, "harder"/"deeper" → tier 3, "max"/"mega"/"ultra" → max. All variants are Opus-only and fire the 🐝 /fswarm* / 🍯 /fmini* statusline indicators.

What This Step Installs

Component	What it does
FidgetFlo CLI + daemon	Coordinates agents, memory, and tasks in the background.
MCP Server	Wires FidgetFlo into Claude Code.
Memory System	Persistent, searchable memory shared across agents + sessions.
Opus Lock	All tasks and spawned agents run on Opus — no silent downgrade to Haiku/Sonnet.
Swarm + Hive + /w4w + /concise + /bullets skills	The commands above.
TypeScript + agentic-flow	Required deps (embeddings, advanced routing).
Statusline	Live indicators for swarms, hives, model, session time, context usage, and Claude rate-limit bars (5h + 7-day).

Want the deep dive? Architecture, agent catalog (60+ types), memory system, hook pipeline, topology options — all in the FidgetFlo repo →

After Step 4

Your core tools are installed. Continue to Step 5 for productivity tools — or open a new cskip session and try something ambitious. FidgetFlo kicks in automatically when the task calls for it.

What's an MCP?

MCP stands for Model Context Protocol — a plugin system for Claude. Connect an MCP once and Claude gets access to a new tool or data source, which it picks up automatically when relevant.

Verify the FidgetFlo MCP is connected

claude mcp list

If FidgetFlo isn't listed, re-add it:

claude mcp add fidgetflo -- npx -y fidgetflo@latest

---

Step 5 - Productivity Tools

Back to top

Connects Claude to the productivity tools you already use. Everything's optional — install only the ones that match your workflow. Once connected they all work through natural language:

"What's on my calendar this week?" · "Add a task called 'Review contracts' due Friday" · "Create a Notion page called Project Roadmap" · "What were the key points from my last meeting?" · "Triage my inbox" · "Trigger my lead-qualification workflow for this email"

Claude picks the right tool automatically based on what you ask. Pick whichever apply:

1. Notion · 2. Granola · 3. n8n · 4. Google Calendar · 5. Morgen ⭐ · 6. Motion Calendar · 7. Playwright · 8. SwiftKit · 9. Superhuman · 10. Google Drive · 11. Vercel

Morgen (5) is the recommended default — it unifies Google, Outlook, iCloud, and native calendars + tasks behind a single API key. Google Calendar (4) and Motion (6) are secondary — install only if you need those accounts directly.

Obsidian MCP lives in 2ndBrain-mogging, not here. Install mogging after this repo completes — it handles vault setup AND registers the Obsidian MCP with Claude Code.

Run Step 5

In a cskip session, paste this:

Important

run this command to install productivity tools: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-5/step-5-install.sh)

The script asks which tools you want, then walks you through each one's credentials. Skip anything you don't use — re-run the script later to add more.

What each tool does

#	Tool	What it does	Needs
1	Notion (@notionhq)	Search, read, create pages + databases in your Notion workspace. Official hosted MCP, 22 tools.	Nothing — browser OAuth on first use. Pick which pages/databases Claude can access during the auth flow.
2	Granola	Search your Granola meeting transcripts + notes through conversation.	Granola installed + signed in on Mac. No key.
3	n8n	HTTP bridge to your own n8n instance — trigger and inspect workflows you built. Not a hosted service.	An n8n workflow with an MCP Server Trigger node; copy its Production URL. Optional Bearer token.
4	Google Calendar	Direct Google Calendar access via OAuth. Secondary — only install if you need a specific Google account bypassing Morgen.	Google account + ~5min to create OAuth creds (script walks you through).
5	Morgen ⭐ 💚	Unified calendar + tasks across Google/Outlook/iCloud/native. Natural-language dates/recurrence, auto-schedule, day reflow. One key for everything.	API key from platform.morgen.so/developers-api.
6	Motion Calendar 💚	Teammate visibility + full event search that Morgen's API doesn't expose. Motion-specific features only.	Motion API key + Firebase key + refresh token + user ID (script walks you through).
7	Playwright (Microsoft)	Lets Claude log into and operate web apps with no API. Runs its own Chromium (not your real browser), reads via accessibility-tree snapshots — fast + reliable.	Node 18+ (from Step 1) + ~hundreds of MB disk for Chromium. No credentials.
8	SwiftKit (swiftkit.sh)	Hosted MCP toolkit for iOS / macOS / Swift development — 100+ tools for writing, building, and shipping Apple-platform code behind one HTTP endpoint. Default for anything iPhone/iOS/Swift-related. Nothing to install locally.	Account + API key (sk_live_ or sk_test_).
9	Superhuman (superhuman.com)	Email triage + drafting from Claude via Superhuman's official remote MCP.	Active Superhuman subscription. One-time browser OAuth on first use.
10	Google Drive	Browse, search, and read Google Drive files — Docs, Sheets, PDFs, shared folders — via Google's official hosted MCP at drivemcp.googleapis.com.	Google account. One-time browser OAuth on first use.
11	Vercel (vercel.com)	Deployments, build logs, runtime logs, domain management, environment variables, and project config via Vercel's official remote MCP.	Vercel account. One-time browser OAuth on first use.

Playwright scope note: Microsoft explicitly says "Playwright MCP is not a security boundary." Treat anything Claude loads through it the same as any browser session you'd drive manually.

💚 = built by fidgetcoding (Morgen, Motion Calendar above; FidgetFlo in Step 4).

After Step 5

Your productivity stack is wired up. Ask about your schedule, add a task, query Notion, trigger a workflow — all from your terminal. Skipped something? Re-run Step 5 later. For Obsidian vault access, install 2ndBrain-mogging.

---

Step 6 - Telegram

Back to top

This step connects Claude to Telegram so you can message it from your phone. You create a bot on Telegram using @BotFather (free, takes about 2 minutes), then the script configures it locally. After setup, you can send Claude messages, photos, and files from any device with Telegram installed. Optional, but it makes Claude accessible from your pocket.

Two launcher commands pair with this step:

• ctg — launches Claude with Telegram connected from any directory. Installed in Step 1. Use this when you want to drive a regular Claude session from your phone.
• cbraintg — same as ctg, but also opens your 2ndBrain vault on launch. Installed by 2ndBrain-mogging. Use this when you want Claude to have vault context while answering Telegram messages.

What It Sets Up

Component	What it does
Telegram Bot	Your personal bot created via @BotFather on Telegram
Bot Token	Stored locally at ~/.claude/channels/telegram/.env
Access Policy	Controls who can message your bot (default: ask before accepting)
ctg command	Launch Claude with Telegram from any directory (installed in Step 1)
cbraintg command	Launch Claude with Telegram inside your 2ndBrain vault (installed by 2ndBrain-mogging)

Run Step 6

Note

Step 6 is interactive — it will ask you to create a bot on Telegram and paste the token. The whole process takes about 2 minutes.

Important

Paste this into your Claude session:

run this command to set up Telegram: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-6/step-6-install.sh)

After Step 6

Open a new terminal and run ctg to launch Claude with Telegram connected. Inside your Claude session, tell it to pair your Telegram bot. Once paired, messages you send from Telegram will appear in your Claude session in real time. You can also use cbraintg to launch with both Telegram and your 2ndBrain vault loaded.

If you see repeating TELEGRAM_BOT_TOKEN required warnings that won't stop: Press Ctrl+C to exit. Your token isn't being detected — Claude Code keeps restarting the Telegram channel in a loop. Use cskip instead of ctg to continue setup, and troubleshoot Telegram separately later. See Troubleshooting → Telegram: stuck in a warning loop.

---

Step 7 - GitHub

Back to top

The GitHub bundle — optional, for devs. Installs four things:

• GitHub CLI (gh) — the terminal binary. Claude shells out to it via Bash for everyday ops (gh pr create, gh issue list, gh repo view). Installs unconditionally — no credentials required. Run gh auth login once after install to sign in.
• GitHub MCP (github/github-mcp-server — GitHub's official hosted server at api.githubcopilot.com/mcp) — Claude gets direct tool-call access to your repos: issues, PRs, files, code search, branches, commits. "List open PRs on cli-maxxing", "search my repos for any file that uses MORGEN_API_KEY" — it just works. Requires a Personal Access Token.
• /gitfix skill — full-repo doc sync. Reads every install script, skill file, and doc, finds drift between code and docs, fixes it. Run it after any significant change so the README stops lying.
• /recon skill — pre-build prior-art recon. Before you build a tool/app/CLI/MCP/library, it sweeps GitHub (via gh) and the web for what already exists, ranks the top ~10 free and paid competitors in a comparison table, finds where an edge actually is (or honestly calls it a red ocean), and ends with a GREEN/YELLOW/RED verdict. Output is a discussion, not code — building starts only after. Auto-offers itself when you say "build / make / start a new X" or "does X exist?"; run it directly with /recon <thing> to skip the offer. No token needed.

Before You Run It

You need a GitHub Personal Access Token (classic PAT) for the MCP. Create one at github.com/settings/tokens/new:

• Name: claude-github-mcp
• Scopes: repo, read:org (under admin:org), gist

Copy the ghp_... value. /gitfix and /recon need no token — they run locally (/recon reads public GitHub through the gh CLI).

Run Step 7

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-7/step-7-install.sh)

Script prompts for your PAT, registers the GitHub MCP (token stored in ~/.claude.json alongside every other MCP credential), and drops /gitfix into ~/.claude/skills/gitfix/ and /recon into ~/.claude/skills/recon/.

What This Step Installs

Component	What it does
GitHub CLI (gh)	Terminal binary. Claude uses it via Bash for PRs, issues, code search, branch ops. Run gh auth login once after install.
GitHub MCP	Exposes GitHub API ops as Claude tools — read/write repos, issues, PRs, code search, branches, commits. Needs a Personal Access Token.
/gitfix skill	Full-repo doc sync. Fixes drift between code and docs. Works on any repo, no token needed.
/recon skill	Pre-build prior-art sweep. Ranks existing free + paid competitors, finds the edge, GREEN/YELLOW/RED verdict. Auto-offers on build-intent. No token needed.

After Step 7

Ask "list my open GitHub issues" or "create a PR on cli-maxxing" and the MCP kicks in automatically. Type /gitfix (or say "sync the repo" / "fix the github" in plain English) after any major change to realign the docs. Run /recon <thing> (or just say "let's build X" and take the offer) before starting anything new — it sweeps what already exists and finds your edge before you write code. To rotate the PAT, re-run Step 7 — it overwrites the token in place.

---

Step 8 - Safety Check

Back to top

Installs /safetycheck — a Claude Code skill that scans any project for security issues. 8 standard API checks + 12 MCP-specific checks auto-activate when it detects an MCP project. Point Claude at a repo, run the command (or say "run a safety check" in plain English), get findings by severity.

It's a first line of defense — the kind of check to run before every deploy, every PR, every handoff. Not a replacement for a full audit.

Run Step 8

In a cskip session, paste:

Important

run this command to install the safety check skill: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-8/step-8-install.sh)

What it checks

API Security (all projects):

• Exposed secrets — API keys, tokens, passwords in source, git history, MCP config.
• Missing rate limiting — unlimited-request endpoints.
• Input sanitization gaps — user input flowing into queries, commands, file paths, MCP tool handlers.
• Dependency vulnerabilities — CVEs in npm/pip packages, MCP SDK version checks.
• Insecure configurations — CORS misconfig, missing .gitignore entries, untracked secrets.

MCP Security (auto-activated on MCP projects):

• Tool description integrity — hidden instructions, injection markers in tool descriptions.
• Unicode smuggling — invisible chars hiding instructions from human reviewers.
• MCP transport security — DNS rebinding, HTTP vs HTTPS, known CVEs (CVE-2025-66414/66416).
• MCP authentication — missing bearer auth on HTTP MCPs.
• Supply chain hygiene — @latest floating versions, rug-pull risk, unverified packages.
• Tool response sanitization — stack traces and raw errors leaking through tool results.
• Audit logging — missing structured logging for tool invocations.

After Step 8

Open any project in Claude and type /safetycheck (or just ask Claude to "run a safety check"). For standard projects, 8 checks run and get reported by severity. For MCP projects, the 12 MCP-specific checks auto-activate.

---

Final Step - Status Line

Back to top

The wrap-up. Installs a custom status line that shows what's active at a glance, plus a final verification pass.

Indicators

Icon	When it shows
⚡️ fidgetflo	FidgetFlo MCP connected
🧠 Brain²	CWD is inside your Obsidian vault (requires 2ndBrain-mogging)
🐝 Swarm	A swarm is active (/fswarm, shows agent count)
🍯 Mini	A mini swarm is active (/fmini, shows agent count)
👑 Hive	A hive-mind is active (/fhive)

The status line also shows your current model, session duration, context-window usage, and Claude rate-limit bars for the 5h and 7-day windows — color-graded green → yellow → red as you approach each limit. Usage is read straight from Claude Code's own rate_limits data (no API token or network call).

Run Final Step

Important

Paste this into your Claude session:

run this command to set up your status line: bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-final/step-final-install.sh)

Verify Everything Works

Once the status line is up, run one last cross-check:

Important

In a fresh cskip session, paste:

Open the cheat sheet at CHEATSHEET.md in this repo and go through every command, skill, and tool listed there. For each one, verify it's installed and working on my machine. If anything is missing, broken, or not configured, fix it. Give me a summary of what passed and what you had to fix.

Claude cross-references CHEATSHEET.md against your actual system, then fixes anything that didn't land — missing skill, unconnected MCP, unregistered alias. Final sanity check.

Manual install + 🧠 Brain² details

Manual install (if you'd rather skip the script):

1. Copy templates/statusline.sh to ~/.claude/statusline.sh
2. Add to ~/.claude/settings.json:

   "statusLine": { "type": "command", "command": "~/.claude/statusline.sh" }

3. Restart Claude Code.

🧠 Brain² indicator: lights up when your CWD is inside the Obsidian vault that 2ndBrain-mogging registered. Mogging's installer writes the vault path to ~/.claude/.mogging-vault; this statusline reads it. No mogging installed → marker doesn't exist → indicator stays hidden (everything else still works). To re-point at a different vault without re-running mogging: echo "$NEW_VAULT" > ~/.claude/.mogging-vault.

Note: Use cskip for this step, not cbrain. The cbrain command requires your Obsidian vault to exist. If you haven't run 2ndBrain-mogging yet, or if something went wrong during vault setup, cbrain will error out. cskip always works.

After Final Step

Setup is complete. Head to You're Ready below for your daily command.

---

Troubleshooting

Back to top

I ran the installer but claude command is not found

The installer adds Homebrew, nvm, and the shell aliases to your config — but those only load when a brand new shell starts. Sourcing the same window often misses.

Fix:

1. Fully close the terminal (Cmd+Q on Mac).
2. Open a fresh terminal.
3. Run:

   claude --version

   You should see something like 2.1.112 (Claude Code).

Still missing? Paste the Reset PATH (stuck install) block from CHEATSHEET.md — it rewires ~/.zshrc with Homebrew, nvm, ~/.local/bin, and the four aliases in one shot.

Some steps say "Homebrew not found" during install

Step 1 installs Homebrew mid-pipeline, but downstream steps in the same shell don't see it yet. Known issue, fixed 2026-04-17.

Fix: close the terminal, open a fresh one, re-run the installer:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/install.sh)

It's idempotent — anything already installed gets skipped.

I see the zsh/bash shell prompt change after install

Modern macOS defaults to zsh even if /etc/passwd still says bash (Terminal.app overrides passwd with its "default login shell" preference). You may notice your prompt looks different after a fresh terminal.

Check which shell you're actually in:

echo $SHELL

Want your passwd entry to match Terminal.app? (Optional — everything works either way.)

chsh -s /bin/zsh

The installer writes to both ~/.zshrc and ~/.bashrc so the aliases work in either shell.

xlsx2csv failed to install

macOS default Python (3.9) ships with PEP 668 restrictions that block pip install into the system Python. The installer now uses pipx to work around this.

Fix: re-run Step 3:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-3/step-3-install.sh)

Still failing? Install manually:

brew install pipx
pipx install xlsx2csv

I want to completely remove everything

See the Uninstall section below — one script reverses the whole stack.

Telegram: pressing Enter skips setup

This is intentional, not a bug. If you hit Enter without pasting a token, Step 6 skips Telegram and continues. Re-run Step 6 later when you have a bot token.

Telegram: stuck in a warning loop after setup

Launching ctg or cbraintg gives a never-ending stream of telegram channel: TELEGRAM_BOT_TOKEN required messages? Your bot token isn't being detected — Claude Code starts the Telegram channel, it exits immediately (no token), Claude Code restarts it, repeat forever.

Fix:

1. Press Ctrl+C to kill the session.
2. Use cskip instead of ctg to keep working — no Telegram needed.
3. Re-run Step 6 to re-enter the token:

   bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-6/step-6-install.sh)

If that doesn't fix it, open cskip and ask Claude:

"My Telegram bot token isn't being detected — can you check my config at ~/.claude/channels/telegram/ and fix it?"

The token file is usually the culprit — missing or wrong format.

Step 5 (Productivity Tools) skips when run through the update command

Step 5 needs interactive input for API credentials. When piped through curl | bash (including the update command), it detects non-interactive mode and exits.

Fix: run Step 5 directly in your terminal:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/step-5/step-5-install.sh)

Obsidian MCP returns internal errors

See the 2ndBrain-mogging troubleshooting guide. The Obsidian MCP is installed and configured by 2ndBrain-mogging.

cbrain says it can't find my vault

See 2ndBrain-mogging — vault setup is handled there. If your vault exists but isn't found, set VAULT_PATH=/path/to/your/vault cbrain.

A step failed or something is missing

Run the update command — it re-runs every step, skips what's already installed, fills in any gaps:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/update.sh)

Or: open a cskip session and describe the problem to Claude. It can diagnose and fix most issues on the spot.

---

You're Ready

Back to top

Quick MCP Check

Before you dive into cbrain, take 30 seconds to verify your MCP connections are live. Open a new terminal and run:

cskip

Once you're inside Claude, type:

/mcp

This shows every MCP server and its connection status. Everything you installed — FidgetFlo, Notion, Granola, n8n, Google Calendar, Morgen, Motion Calendar, Playwright, SwiftKit, Superhuman, Google Drive, Vercel, GitHub (if you ran Step 7), Obsidian (from 2ndBrain-mogging), and design tools (from creativity-maxxing) — should show as Connected. If anything shows as failed or disconnected, just tell Claude:

"One of my MCP servers isn't connecting — can you troubleshoot it?"

Claude will diagnose and fix it on the spot. Once everything is green, you're ready.

---

Everything is installed, configured, and wired together. From now on, this is the only command you need:

Important

Your daily command:

cbrain

That's it. cbrain opens Claude Code directly inside your 2ndBrain vault with all permissions skipped. Your vault is your home base — every tool, skill, and MCP server you just installed is available the moment you type it.

Haven't run 2ndBrain-mogging yet? Use cskip instead of cbrain until your Second Brain vault is set up. cbrain requires the Obsidian vault to exist — it will error if you haven't created one. Once 2ndBrain-mogging is complete, switch to cbrain as your daily driver.

What cbrain gives you:

• Drops you into your Obsidian vault automatically — no cd-ing around
• All permissions skipped — Claude acts immediately, no approval prompts
• Full access to everything: /fswarm (+ tiers 1/2/3/max), /fmini (+ tiers 1/2/3/max), /fhive, /w4w, /concise, /bullets, /safetycheck, /gitfix, /recon, FidgetFlo, Notion, Granola, n8n, Google Calendar, Morgen, Motion Calendar, Playwright, SwiftKit, Superhuman, Google Drive, Vercel, Obsidian, design tools, video tools — all of it
• Your status line shows what's active at a glance

When to use something else:

• cskip — when you need to work outside your vault (a different project, a random folder)
• cbraintg — same as cbrain but with your Telegram channel connected
• ctg — skip-permissions + Telegram from any directory

But for day-to-day use? Just type cbrain and go.

---

Installation Order

Back to top

Run the steps in this order:

Step	Name	What it does
1	CLI Tools	Git, Node.js, Claude Code, shell aliases
2	Bonus Software	Ghostty (GPU-accelerated terminal) + Arc (power-user browser). Optional but recommended.
3	Developer & Utility Tools	Python, Pandoc, jq, ripgrep, no-flicker mode, etc.
4	FidgetFlo	Multi-agent orchestration — swarms, hives, persistent memory, Opus-locked
5	Productivity Tools	Notion + Granola + n8n + Google Calendar + Morgen + Motion Calendar + Playwright + SwiftKit + Superhuman + Google Drive + Vercel (all optional — pick what you use; Morgen recommended)
6	Telegram	Telegram bot setup — message Claude from your phone. Press Enter to skip if you don't have a bot yet.
7	GitHub	GitHub CLI (gh) + GitHub MCP (repos, issues, PRs, code search — MCP requires PAT) + /gitfix skill for full-repo doc sync + /recon skill for pre-build prior-art recon
8	Safety Check	Security auditing — 8 API checks + 12 MCP checks for tool poisoning, DNS rebinding, supply chain attacks
Final	Status Line	Status indicators + system health check

Note: Step 5 (Productivity Tools) is all optional — install only the tools you use. Step 6 (Telegram) is optional — press Enter to skip if you don't have a bot token yet; you can always re-run it later. Step 7 (GitHub) is optional — skip it if you don't use GitHub with Claude. Step 8 (Safety Check) installs a security auditing skill — 8 standard checks for any project, plus 12 MCP-specific checks that auto-activate when an MCP project is detected. The Final Step (Status Line) is the wrap-up — it wires your status indicators and runs a system health check.

---

Staying Up to Date

Back to top

This command re-runs every step, skips anything already installed, and picks up anything new. It covers everything in this repo as of right now. If new steps get added in the future, the update command will include them too.

Open your terminal and run cskip to start a Claude session, then paste the update command. Or if you prefer, just paste it directly into your terminal without Claude.

Important

Paste this into your terminal:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/update.sh)

---

Uninstall

Back to top

One script reverses the whole stack. Your Obsidian vault, notes, and Claude account are never touched.

Important

Paste this into your terminal:

bash <(curl -fsSL https://raw.githubusercontent.com/fidgetcoding/cli-maxxing/main/uninstall.sh)

Removes the cli-maxxing aliases (cskip, cc, ccr, ccc), the ctg script, all MCPs this setup installed, all FidgetFlo skills + /w4w + /concise + /bullets + /safetycheck + /gitfix + /recon, dev tools, Arc Browser, and the Ghostty config. cbrain and cbraintg are managed by 2ndBrain-mogging and are not touched here. The YouTube / Instagram transcription stack (yt-dlp, whisper-mcp, ffmpeg, Whisper models) lives in creativity-maxxing — run its uninstaller separately if you installed it.

Keeps: Homebrew, Git, Node.js, Claude Code itself, your Obsidian vault + notes, your Claude account — general-purpose tools + your data. The script prints manual-removal commands at the end if you want a fully clean machine.

Full list of what gets removed

• Claude Code shell aliases (cskip, cc, ccr, ccc) and the ctg script (~/.local/bin/ctg). cbrain and cbraintg are managed by 2ndBrain-mogging — not removed here.
• All MCPs installed by this repo: FidgetFlo, Notion, Granola, n8n, Google Calendar, Morgen, Motion Calendar, Playwright, SwiftKit, Superhuman, Google Drive, GitHub — design + media MCPs are managed by creativity-maxxing; Obsidian is managed by 2ndBrain-mogging
• All skills: fswarm*, fmini*, fhive, w4w, concise, bullets, gitfix, recon, safetycheck — UI/UX Pro Max + Taste Skill pack + Remotion are managed by creativity-maxxing
• Dev tools: pandoc, jq, ripgrep, tree, fzf, wget, weasyprint, ffmpeg, xlsx2csv, poppler
• GitHub CLI (gh — installed by Step 7 alongside the GitHub MCP + /gitfix + /recon skills)
• Motion Calendar config (~/.motion-mcp/)
• Google Calendar config (~/.google-calendar-mcp/)
• Arc Browser (if installed via Step 2)
• Ghostty config (the app itself is kept)
• Status line config

---

More Coming Soon

Back to top

This setup is a living project. New steps, tools, and workflows will be added as they're ready. If you have the update command above, you'll always be able to catch up with one paste.

---

License

MIT — see LICENSE.

---

Built by Nate Davidovich / Lorecraft

⤴ back to top

Security: gitleaks pre-commit hook

This repo ships with a .gitleaks.toml config and a one-liner installer for a local pre-commit hook that scans staged content for secrets (GitHub tokens, API keys, JWTs, etc.) before every commit.

bash scripts/install-pre-commit-hook.sh

The hook runs gitleaks protect --staged and blocks commits that contain secrets. For emergencies you can bypass with git commit --no-verify — but DO NOT bypass for real secrets. Use env vars or a secret manager instead.

If gitleaks isn't installed yet:

• macOS: brew install gitleaks
• Linux: https://github.com/gitleaks/gitleaks/releases