---

name: ec2 description: Manage AWS EC2 instances for GPU workloads and Spark jobs. Use this skill to start, stop, or check status of EC2 instances. Invoke with /ec2.

EC2 Instance Management

This skill manages EC2 instances for GPU workloads and Spark jobs using aws/ec2_control.py.

Available Commands

List Instances

Copypython aws/ec2_control.py list

Detailed Status

Copypython aws/ec2_control.py status

Shows instance state, public/private IPs, and instance types.

Start Instance

Copy# By index (from list command)
python aws/ec2_control.py start --select 1

# By name
python aws/ec2_control.py start --name simd_gpu

Stop Instance

Copypython aws/ec2_control.py stop --select 1

Restart Instance

Copypython aws/ec2_control.py restart --select 1

Predefined Instances

Regions

Supports multiple regions:

• eu-west-2 (default)
• us-east-1
• us-west-2

Copypython aws/ec2_control.py list --region us-east-1

SSO Login

The script handles AWS SSO login automatically. If session expired, it will prompt for re-authentication.

Common Workflows

Start GPU Instance for VLM Embedding

Copy# Start GPU instance
python aws/ec2_control.py start --name simd_gpu

# Wait for instance to be ready, then SSH
ssh simd_gpu

# Start embedding service
screen -S vlm
python -m wc_simd.vlm_embed_service

Remote Development

Copy# Start instance
python aws/ec2_control.py start --select 1

# Get IP and connect via VS Code Remote SSH
python aws/ec2_control.py status