The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill contains numerous AWS CLI and boto3 examples for managing cloud infrastructure. All commands are consistent with the stated purpose of EC2 management.
[DATA_EXPOSURE] (LOW): The skill references local file paths like ~/.ssh/id_rsa.pub and creates my-key.pem. These are standard operations for managing SSH access to virtual machines and do not involve exfiltration.
[OBFUSCATION] (SAFE): The UserData field in the launch template example uses Base64 encoding (IyEvYmluL2Jhc2gKeXVtIHVwZGF0ZSAteQo=). This is the standard format required by the AWS API for this field. The content decodes to a benign script: #!/bin/bash\nyum update -y.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill has a data ingestion surface as it reads status and metadata from AWS resources (e.g., via describe-instances). While this is a theoretical attack vector if an attacker controls resource names or tags, the skill follows best practices and does not perform unsafe execution based on this data.

ec2