27 Claude Code TIPS you didn't know about (in 12 minutes)

Concise summary — 27 Claude Code tips

This is a cleaned, organized Markdown version of the 27 tips for using Claude Code (technologies, features, workflows, and guides). Tips are grouped for readability: setup, workspace/workflow, agent configuration, project architecture, controls, automation, and resources.

Setup & core tools

• Install a code editor: VS Code or Anti-Gravity (both free).
• Add the Claude Code plugin to integrate Claude into your code workspace.
• Use Dribbble for visual inspiration; attach screenshots or URLs for pixel-for-pixel cloning.
• Import free 3D assets (e.g., from Spline) for interactive visuals and cursor-following elements.
• Use built-in OS dictation to speak prompts (mac: FN double-tap).

Workspace & multitasking

• Open multiple tabs/conversations and use split-screen (drag conversations) to run parallel tasks.
• Message queuing: send multiple messages without waiting—Claude will queue and process them sequentially.
• Delegate sub-tasks to parallel agents (sub-agents) for faster, specialized outputs and cleaner context separation.
• Interrupt & rewind: press Escape to interrupt runaway generations; use rewind to revert to a prior message/state.
• Enable auto-save in VS Code / Anti-Gravity to avoid losing edits.

Agent configuration — required project files

Always keep these two living files in each project:

1. Claude.md — defines agent behavior and training.
2. project_specs — describes what to build.

Treat both as living documents that the agent always reads.

What to include in Claude.md

• Have Claude challenge your assumptions rather than act as a “yes-man.”
• Add a quality gate: require honest scoring and actionable improvements before approving outputs.
• Require automated testing (unit/integration checks) before delivering a final product.
• Preserve high-value context and avoid token waste; ensure conversation compaction doesn’t lose critical info.
• Include “upgrade suggestions” so Claude proposes continuous improvements after each interaction.
• Use Claude.md to define personal/project defaults and persistent memory (editable).

Project architecture & reusable components

Follow a three-layer architecture:

• Layer 1: Claude.md — always read; defines agent behavior.
• Layer 2: Workflows / Skills — on-demand automations (reusable tasks like “reply-to-email”).
• Layer 3: Reference files — single-source-of-truth assets (tone, brand voice, templates) referenced by skills.

Additional points:

• Build reusable skills to avoid re-providing context and to improve iteratively.
• Keep shared reference files so updates propagate across skills and projects.

Automation, plugins & research

• Use /plugin to access/install pre-built solutions (e.g., front-end design plugins) to accelerate work.
• Use an auto-research package to run tests and iterative improvements autonomously without manual involvement.
• Require automated tests in the delivery pipeline (see Claude.md notes).

Response structure and behavior expectations

Force Claude to respond in a predictable format. A recommended structure:

What it did, what it needs from you, why it matters (simple explanation), next steps, and any errors/context.

This makes outputs actionable and easy to review.

Permissions, compaction & context management

• Toggle “bypass permissions” in Configure Editor to prevent Claude from asking for manual approvals mid-workflow.
• Monitor the conversation/context window: compaction occurs when overloaded. Compact proactively and add reminder notes for details you must not lose.
• Use the persistent memory feature for personal and project defaults; edit memory as needed.

Productivity tips & editor controls

• Dictation: use OS-level dictation to speak prompts instead of typing.
• Message queuing: queue multiple messages; Claude processes sequentially.
• Interrupt long or runaway outputs with Escape; use rewind to revert to a prior state.
• Auto-save in your code editor to prevent lost work.

Insights & analytics

• Type “insights” to get a report of lifetime usage, projects, errors, and behavior analytics from Claude.

Free resources, training & services

The creator referenced several resources and offerings:

• Free Claude.md template (description/link referenced).
• Free 4.5-hour Claude Code masterclass on YouTube.
• Free community with blueprints.
• Paid community offering training to build AI automation agencies or automate an existing business.
• Agency service to implement solutions for clients.

Main speakers / referenced tools

• Main speaker: the video creator / narrator (owner of the Claude Code masterclass and communities).
• Tools/sources referenced: Claude Code (Anthropic agent workspace), VS Code, Anti-Gravity, Dribbble, Spline, built-in OS dictation, installed plugins (front-end design, auto-research package).