Installer

Katana Installer — Agent Setup Skill

Overview

This skill turns any CLI agent into a Katana setup wizard. When triggered, walk the user through configuring their Katana Agent — API keys, user profile, skill setup, and preferences.

How to Start

Two entry points:

1. CLI bootstrap first: katana memory init — ensures the ~/.katana/ data folder exists before continuing with setup
2. Direct: User says /setup or "set up my katana" — you handle everything

Step 1: Audit Current State

Read these files before doing anything:

ls ~/.katana/commands/
ls ~/.katana/memory/skills/
cat ~/.katana/memory/core/user.md
cat ~/.katana/settings.json

Report what you find: how many commands, how many skills, whether user.md is filled in, what's configured vs needs setup.

Step 2: User Profile

If ~/.katana/memory/core/user.md is empty or minimal, ask conversationally:

• What's your name?
• Where are you located? (timezone)
• What are you building / working on?
• How do you like your agent to communicate?
• Anything else the agent should always know?

Write answers to user.md. Have a natural conversation, not a form.

Step 3: Walk Through Skills Needing Configuration

Skills That Need API Keys

Skill	What's Needed	Where to Get It
x-auto-dm	X API key + OAuth tokens	https://developer.x.com/en/portal/dashboard
x-thread	X API key (same as above)	Same X developer portal
telegram	Telegram bot token	Message @BotFather on Telegram
reddit-bot	Reddit client ID + secret	https://www.reddit.com/prefs/apps
google-workspace-gog	Google OAuth credentials	https://console.cloud.google.com/apis/credentials
go-places	Google Places API key	https://console.cloud.google.com/apis/credentials
threads-bot	Meta/Threads API access	https://developers.facebook.com

Skills That Need Software Installed

Skill	Dependency	Install Command
playwright	Playwright browsers	npm install -g playwright && npx playwright install
email-non-gmail-himalaya	Himalaya CLI	brew install himalaya or cargo install himalaya

Skills Ready Out of Box (No Setup)

obsidian, obsidian-memory, remember, recall, email-best-practices, weather, trello, new-skill, new-skill-builder, new-mcp-builder, stream, pulse, competitive-ads-extractor, RentAHuman, and all agent skills (oracle, katana, vibe-curator).

For Each Skill That Needs Config

1. Explain what it does in 1 sentence
2. Ask "Want to set this up?"
3. If yes: provide the URL, ask for the key, write to .env in the skill folder
4. If no: skip

Writing API Keys

# X API keys → x-auto-dm skill folder
echo "X_BEARER_TOKEN=your_token" > ~/.katana/memory/skills/social/x-auto-dm/.env

# Telegram bot token
echo "TELEGRAM_BOT_TOKEN=your_token" > ~/.katana/memory/skills/social/telegram/.env

# Google Places
echo "GOOGLE_PLACES_API_KEY=your_key" > ~/.katana/memory/skills/productivity/go-places/.env

NEVER ask the user to manually edit files. Offer to write values for them.

Step 4: Settings.json

Show current permissions from ~/.katana/settings.json. Explain what each does. Ask if they want to add MCP servers or permissions.

Step 5: Summary

⚡ Katana Setup Complete

Profile:
  ✅ user.md configured

Skills Configured:
  ✅ x-auto-dm — X API connected
  ✅ telegram — @YourBot linked
  ⏭️ reddit-bot — skipped
  ✅ playwright — installed
  ✅ stream — 13 RSS feeds ready

Ready to Use:
  /katana — master agent (morning routine, daily tasks)
  /oracle — astrology advisor
  /stream — news & market briefing
  /vibe-curator — lifestyle agent

Next:
  • cd into any project → katana claude init
  • /stream for your first news briefing
  • /remember at end of sessions to save context

Rules

• NEVER display API keys back to the user
• NEVER store keys in memory files (soul.md, user.md, work.md)
• Keys go ONLY in .env files inside skill folders
• Be conversational, not robotic
• Let users skip anything — they can come back with /setup