claude code #ai

README.MD

installation

• curl -fsSL https://claude.ai/install.sh | bash
• claude --version

load skills

• mkdir ~/.claude
• ln -s /var/www/boilerplate/AGENTS.md ~/.claude/CLAUDE.md
• ln -s /var/www/boilerplate/SKILLS ~/.claude/skills

usage

• IS_SANDBOX=1 claude --continue --dangerously-skip-permissions --effort high
• claude
• claude --continue
• /login
• /model
• /usage
• /compact
• /permissions
• ctrl+enter

settings

• nano ~/.claude/settings.json

{
  "skipDangerousModePermissionPrompt": true,
  "spinnerTipsEnabled": false,
  "hooks": {
    "Stop": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "bash /mnt/o/DOCS/SCRIPTS/NOTIFY/notify.sh claude-stop",
            "async": true,
            "timeout": 20
          }
        ]
      }
    ]
  }
}

agent teams

• ~/.claude/settings.json
• { "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" } }
• apt-get install -y tmux
• echo "set -g mouse on" >> ~/.tmux.conf
• tmux new-session -s claude
• cd /var/www/my-project
• nvm use --lts
• git init
• touch CLAUDE.md
• git add -A . && git commit -m "init"
• claude
• exit
• tmux attach -t claude
• tmux kill-session -t claude

prompt

Erstelle ein Snake-Spiel als Webapplikation mit:
- Frontend: HTML5 Canvas Game (index.html, game.js, styles.css)
- Backend: Node.js Express Server mit Highscore-API (server.js, routes/, db/)
- Tests: Unit-Tests für Game-Logik und API-Endpoints

Phase 1: Plane die Architektur in CLAUDE.md mit klarer Task-Aufteilung.
Phase 2: Erstelle ein Team — ein Teammate für Frontend, einer für Backend, 
einer für Tests. Alle arbeiten parallel.

ralph mode

install

• mkdir /tmp/ralph-claude-code
• cd /tmp/ralph-claude-code
• git clone https://github.com/frankbria/ralph-claude-code.git .
• ./install.sh
• cd ~/
• rm -rf /tmp/ralph-claude-code

usage

• cd my-project
• init: ralph-enable
• run: ralph --monitor
• stop: pkill -9 -f 'ralph_loop.sh|claude --model|timeout.*claude'; tmux ls 2>/dev/null | cut -d: -f1 | xargs -rI{} tmux kill-session -t {}
• log: tail -F -q .ralph/logs/claude_output_*.log | jq .
• reset (soft): find .ralph -mindepth 1 -maxdepth 1 ! -name '*.md' ! -name '.gitignore' -exec rm -rf {} +
• reset (hard): rm -rf .ralph .ralphrc && ralph-enable
• show all sessions: claude --resume > CTRL+A

.ralphrc

CLAUDE_MODEL="claude-opus-4-7"
ALLOWED_TOOLS="Write,Read,Edit,Grep,Glob,Bash,Monitor,WebFetch,WebSearch,Task,TodoWrite,NotebookEdit"
MAX_CALLS_PER_HOUR=30
MAX_TOKENS_PER_HOUR=800000
CB_NO_PROGRESS_THRESHOLD=2

gitignore

option 1

/.ralph
/.ralphrc

option 2

/.ralph/*
!/.ralph/.gitignore
!/.ralph/*.md

security pentest

installation

• cd /var/www/project
• claude
• /plugin marketplace add trailofbits/skills
• /plugin marketplace add anthropics/claude-plugins-official
• /plugin install static-analysis@trailofbits
• /plugin install insecure-defaults@trailofbits
• /plugin install sharp-edges@trailofbits
• /plugin install supply-chain-risk-auditor@trailofbits
• /plugin install differential-review@trailofbits
• /plugin install audit-context-building@trailofbits
• /plugin install variant-analysis@trailofbits

usage

• cd /var/www/project
• claude
• security-review against http://localhost:8080

mcp

• claude mcp add playwright npx @playwright/mcp@latest
• alternative: claude code chrome (currently not supported in wsl): https://code.claude.com/docs/de/chrome
• claude mcp add context7 -- npx -y @upstash/context7-mcp