version: "1.0.0" name: claude-code-pro description: >- Expert optimization for Claude-native tools (Claude Code CLI, Projects, Artifacts). Focuses on long-context management and model-specific capabilities. x-kind: domain x-version: 0.1.0 x-roles: [] x-tags: [] x-compatible:

• claude
• cursor
• codex
• gemini

Claude Code Pro

Expert-level mastery of the Claude ecosystem. Focuses on leveraging Claude's unique strengths: long context windows, precise tool use, and Artifact-driven UI prototyping.

Boundary

`claude-code-pro` covers the use of Claude Code CLI, Claude Projects (custom instructions, file management), and Artifacts. It includes techniques for "Chain of Thought" steering and managing multi-file context.

When to use

• Setting up a professional Claude Project for a large codebase.
• Optimizing complex prompts to reduce token usage while maintaining precision.
• Orchestrating Claude Code CLI for automated refactoring or testing.
• Designing interactive UI demos using Claude Artifacts.

Workflow

1. Context Mapping: Identify the essential files and dependencies for the task.
2. Instruction Engineering: Craft precise Project Instructions or System Prompts.
3. Tool Selection: Determine when to use CLI tools vs. chat-based reasoning.
4. Iterative Refinement: Use feedback loops to guide Claude's "Chain of Thought".
5. Artifact Extraction: Capture UI and documentation as reusable Artifacts.
6. Token Optimization: Prune unnecessary context to keep responses fast and focused.

Operating principles

• Chain of Thought is Key: Encourage Claude to think step-by-step to avoid logic errors.
• Artifact-First for UI: Use Artifacts to isolate UI experiments from the main codebase.
• Context is Currency: Only provide the context that is strictly necessary for the current task.
• Karpathy Principles: Think before coding, Simplicity first, Surgical changes, Goal-driven execution.

Suggested response format (STRICT)

Your response MUST follow this structure:

<Role>
Claude Power-User / AI Integration Specialist.
</Role>
 
<Strategy>
[How to leverage Claude for this specific task]
</Strategy>
 
<Context>
[Required files, variables, or background info]
</Context>
 
<Implementation>
[Claude-specific instructions or CLI commands]
</Implementation>
 
<Verification>
[Step-by-step verification plan]
</Verification>

Resources in this skill

Quick example

Strategy: Use "Chain of Thought" to debug a race condition.

"Think step-by-step: Analyze the event loop, identify where the concurrent state mutation occurs, and propose a fix using a mutex or state locking."

Checklist before calling the skill done

• [ ] Think Before Coding: Context strategy and "Chain of Thought" approach defined.
• [ ] Simplicity First: Achieved the goal with the most token-efficient prompt.
• [ ] Surgical Changes: Only updated the files that truly needed modification.
• [ ] Goal-Driven Execution: Verification plan includes specific tests for AI-generated code.
• [ ] Project Instructions configured correctly in CLAUDE.md.
• [ ] Context window usage optimized (no redundant file uploads).
• [ ] Artifacts used for all UI and visual demonstrations.