Skip to content

Instantly share code, notes, and snippets.

@malpern
Last active June 25, 2025 03:23
Show Gist options
  • Select an option

  • Save malpern/50818345b562e8be6dc11013c2579886 to your computer and use it in GitHub Desktop.

Select an option

Save malpern/50818345b562e8be6dc11013c2579886 to your computer and use it in GitHub Desktop.
Hacker Dojo: Python Meet up: June 2025

Micah Demo

https://github.com/malpern/KeyPath

Claude Code

image
npm install -g @anthropic-ai/claude-code

Claude Code: A New Kind of AI Assistant

  1. Terminal-based, not IDE
  2. Works with all tools
  3. Fits into all existing workflows
  4. General purpose – you can use it for almost anything
  5. Infinitely hackable
image

vs Cursor

  1. IDE vs Terminal
  2. $20 vs pay as you go OR, better value, Claude Pro ($100 a month) and Claude Max ($200 a month).
image
  1. Installation & Initial Setup
    • Installing Claude Code (e.g., npm install -g @anthropic-ai/claude-code).
    • Essential first setup commands:
      • /terminal-setup: For better multiline input (Shift+Enter).
      • /theme: Customizing light/dark mode.
      • /config: Managing notifications.
      • /allowed-tools: Setting tool permissions.
      • /install-github-app: Integrating with GitHub issues and PRs.
    • Leveraging MacOS dictation for easier prompting.

Getting Started & Core Concepts

image
  1. Codebase Q&A: Your First Interaction
    • Why it's the easiest way to start.
    • No initial setup beyond installation.
    • Data privacy: Your code stays local.
    • Example Q&A Prompts:
      • Understanding code usage (e.g., "How is @RoutingController.py used?").
      • Investigating code history (e.g., "Why does recoverFromException take so many arguments? Look through git history to answer.").
image
  1. Effective Prompting Basics (Tips #1 & #2)
    • Start with codebase Q&A to understand Claude's capabilities.
    • Practice prompting to learn what Claude "gets" intuitively versus what requires more specific instructions.

Working with Code & Tools

image
  1. Editing Code with Claude

    • Moving beyond Q&A to active code modification.
  2. Leveraging Built-in Tools

    • Overview of Claude Code's out-of-the-box tools (bash, file search, file listing, read/write, web fetch, TODOs, sub-agents).
    • How tools make Claude Code powerful.
image
  1. Steering Claude & Advanced Prompting

    • Guiding Claude to use tools effectively.
    • Example Prompts for Complex Tasks:
      • Proposing and implementing fixes (e.g., "Propose a few fixes for issue #8732, then implement the one I pick.").
      • In-depth analysis (e.g., "Identify edge cases... think hard.").
      • Automating workflows (e.g., "commit, push, pr").
      • Using parallel agents for brainstorming.
    • Strategy: Ask Claude to plan or brainstorm before generating code.
  2. Integrating Your Team's Custom Tools (Tip #3)

    • Teaching Claude about your specific bash scripts or custom CLI tools (MCPs).
    • Example: "Use the barley CLI to check for error logs... Use -h to check how to use it."
    • Adding MCP tools (e.g., claude mcp add barley_server -- node myserver).
image

Module 4: Advanced Techniques & Workflows

  1. Common Development Workflows with Claude (Tip #4)
    • Exploration & Bug Fixing: Explore > Plan > Confirm > Code > Commit.
      • Example: "Figure out the root cause for issue #983, then propose a few fixes. Let me choose an approach before you code."
    • Test-Driven Development: Write Tests > Commit > Code > Iterate > Commit.
      • Example: "Write tests for @utils/markdown.ts to make sure links render properly..."
    • UI Implementation & Iteration: Write Code > Screenshot Result > Iterate.
      • Example: "Implement [mock.png]. Then screenshot it with Puppeteer and iterate till it looks like the mock."
image
  1. Maximizing Performance with Context (Tips #5 & #6)
    • The more context, the smarter Claude becomes.
    • Methods for providing context:
      • CLAUDE.md files:
        • User home (~/.claude/CLAUDE.md, shared across all user projects).
        • Project root (PROJECT_ROOT/CLAUDE.md, checked in).
        • Local project (PROJECT_ROOT/CLAUDE.local.md, not checked in).
      • Slash commands (e.g., /user:foo from ~/claude/commands/foo.md).
      • At-mentioning files/folders (e.g., @a/foo.py).
    • How context is pulled in (automatically vs. on-demand).
    • Importance of tuning context:
      • Is it for you or the whole team?
      • Should it be added automatically or lazily?

Module 5: Power User Features & Automation

image
  1. Essential Keybindings
    • Shift+Tab: Auto-accept edits.
    • #: Create a memory.
    • !: Enter bash mode.
    • @: Add a file/folder to context.
    • Esc: Cancel current operation.
    • Double-Esc: Jump back in history (--resume to continue).
    • Ctrl+R: View verbose output.
    • /vibe (briefly mentioned for changing Claude's style).
image image
  1. Scripting Claude: The Claude Code SDK
    • Programmatic, low-level access.
    • Use cases: CI/CD, non-interactive automation, building custom interactive apps.
    • Current support: CLI (TypeScript). Python SDK upcoming.
    • Example CLI usage: claude -p "prompt" --allowedTools Tool(params) --output-format json.
    • Using Claude Code as a Unix utility (piping input/output).
      • Example: git status | claude -p "what are my changes?" --output-format=json | jq '.result'
image
  1. Multi-Claude: Running Claudes in Parallel
    • Techniques for parallel work:
      • Multiple terminal tabs with different checkouts.
      • Single checkout with git worktrees.
      • SSH + TMUX sessions.
      • GitHub Actions for parallel job execution.

An interesting essay: https://steipete.me/posts/2025/claude-code-is-my-computer

image

Q&A and Advanced Insights

  1. Implementation Challenges: Insights into building Claude Code (e.g., Bash safety, permissions).
  2. Multimodal Capabilities: Using images with Claude Code (drag & drop, file paths, copy-paste).
  3. CLI vs. IDE: Rationale behind the CLI-first approach.

This outline provides a structured way to present the information from the video, highlighting key commands, concepts, and best practices for using Claude Code effectively.

Micah Tips: 0. If you have questions about how to use it, just ask!

  1. Need a good terminal app. eg. Mac Ghosty.
  2. @ files
  3. claude --dangerously-skip-permissions, just have backups or run in docker, or just use shift-tab
  4. claude --continue
  5. Context compacting
  6. Drag and drop of control v paste images
  7. Tell it to think hard, or ultrathink if you want to burn more tokens
  8. Planning mode or just ask it to plan (or write planning files)
  9. MCP: Eg Puppeitier, Context7. You can just ask it to install these and then restart claude code.
  10. Claude.md files
  11. Peter Steinberger
  12. https://github.com/steipete/agent-rules
  13. https://github.com/steipete/macos-automator-mcp
  14. https://github.com/steipete/Peekaboo
  15. Claude Code SDK
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment