UpdatesConferences

Claude Code First Meetup Oslo

Oslo, Norway

Skills & Plugins

Context Management

  • Context is king - Less context means AI Slop
  • Keep the eye on the context
  • Use ccusage statusline to monitor
  • Summarize, inject new MD file, do /clear
  • Make a habit of context hygiene
  • Create slash commands for common workflows
  • Docs for coding with AI - example: Supabase

Claude MD & Slash Commands

  • CLAUDE.md serves as your AI knowledge base
  • /init command to set up
  • Slash commands with arguments supported
  • In CLAUDE.md you can reference slash commands!
  • Include bash commands in slash commands
  • Can use # for memories

Sub Agents

  • Have their own context window
  • Can have indefinite context
  • Without cluttering main context window
  • Chain subagents together for complex workflows

Prompting Tips

  • Prompt "Make it better" to trigger improvement behavior and force LLM for ideas
  • Use ultra think mode for complex reasoning
  • Use plan mode for structured approach
  • Put plan in MD file to keep track and reuse

Keyboard Shortcuts

  • Shift + Tab and Tab only for navigation

Hooks

  • Guardrails - stop LLM from doing something
  • Create a hook to send a Slack message to tell agent to continue

MCP (Model Context Protocol)

  • Eating your context all the time
  • Tool search tool from Anthropic to overcome - check it out

GitHub Integration

  • Can be done from GitHub app, cloud, sandbox
  • Write - PR and review workflow
  • New Slack integration available
  • Use GitHub issues to track Claude ideas

Code Quality

  • Be owner of the code - Use code reviews. Take help from AI agents
  • Human in the loop - Always review AI output
  • OTEL integration with Claude Code
  • Record frontend with 3 agents, then have them agree
  • Silent failure mode punchers - catches many things
  • Review agents for quality
  • An agent obsessed with silent failures and corner cases
  • Biggest issue in AI Slop: works locally but not globally