Skip to main content

​
Skill Installation

Wafer provides AI coding assistant skills that integrate with Claude Code, Cursor, Codex, and other tools. These skills give AI assistants knowledge of Wafer commands and GPU development best practices.

​
Quick Start

# Check current status
wafer skill status

# Install the skill
wafer skill install

# Uninstall if needed
wafer skill uninstall

​
Commands

​
wafer skill install

Install the Wafer skill for AI assistants: 
wafer skill install [OPTIONS]
Options:
OptionDescription
--targetTarget assistant: claude-code, cursor, codex, all
--forceOverwrite existing installation
Examples: 
# Install for all supported assistants
wafer skill install

# Install for specific assistant
wafer skill install --target claude-code

# Force reinstall
wafer skill install --force

​
wafer skill uninstall

Remove the Wafer skill: 
wafer skill uninstall [OPTIONS]
Options:
OptionDescription
--targetTarget assistant to uninstall from

​
wafer skill status

Check installation status: 
wafer skill status
Output: 
Wafer Skill Status:

  Claude Code    installed (v1.2.3)
                 Path: ~/.claude/skills/wafer-guide/

  Cursor         installed (v1.2.3)
                 Path: ~/.cursor/skills/wafer-guide/

  Codex          not installed

​
Supported Assistants

​
Claude Code

The skill integrates with Claude Code’s skill system: 
wafer skill install --target claude-code
After installation, Claude Code can:
  • Run Wafer CLI commands
  • Understand GPU profiling output
  • Suggest kernel optimizations
  • Navigate Wafer documentation

​
Cursor

Install for Cursor IDE: 
wafer skill install --target cursor
The skill adds Wafer context to Cursor’s AI features.

​
Codex / Other

For other assistants: 
wafer skill install --target codex

​
What the Skill Provides

The Wafer skill includes:
  • Command Reference: Full documentation of Wafer CLI commands
  • Best Practices: GPU kernel development patterns
  • Profiling Guidance: How to interpret NCU, nsys, and other tool output
  • Optimization Strategies: Common optimization techniques for CUDA/Triton/HIP

​
Using with Claude Code

After installing the skill, use Wafer commands naturally: 
You: Profile my kernel and find the bottleneck

Claude: I'll run wafer nvidia ncu analyze on your kernel...
The skill enables Claude to:
  • Understand Wafer-specific concepts
  • Suggest appropriate Wafer commands
  • Interpret profiling results
  • Guide optimization workflows

​
Updating the Skill

Update to the latest version: 
wafer skill install --force
This reinstalls the skill with the latest content.

​
Skill Location

Skills are installed in:
AssistantLocation
Claude Code~/.claude/skills/wafer-guide/
Cursor~/.cursor/skills/wafer-guide/
Codex~/.codex/skills/wafer-guide/

​
Troubleshooting

Restart the AI assistant after installation:
  • Claude Code: Close and reopen terminal
  • Cursor: Reload window (Cmd/Ctrl + Shift + P → “Reload Window”)
Check write permissions for the skill directory:
ls -la ~/.claude/skills/
Force reinstall:
wafer skill install --force

​
Next Steps

AI Agent

Use Wafer’s built-in AI agent.

Demo

Try interactive demos.

Quickstart

Get started with Wafer.

CLI Overview

Learn CLI basics.