The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands for interacting with cloud infrastructure using specialized CLI tools for Azure, AWS, and GCP.
[PRIVILEGE_ESCALATION]: Includes instructions for elevating permissions, such as using sudo for tool installation and assigning highly privileged roles like 'Global Administrator' in Azure tenants as seen in references/advanced-cloud-scripts.md.
[PERSISTENCE]: Documents methods for maintaining long-term access, including creating backdoor service principals, additional IAM access keys, and new administrative users in references/detailed-guide.md.
[DATA_EXFILTRATION]: Contains commands designed to extract sensitive data, including syncing S3 buckets, dumping secrets from Azure Key Vaults, and retrieving environment variables from serverless functions.
[CREDENTIALS_UNSAFE]: Provides scripts to search for passwords in user attributes and instructions to retrieve identity tokens from cloud metadata services (IMDS).
[EXTERNAL_DOWNLOADS]: Fetches installation scripts and binaries for AWS and GCP CLI tools from official, well-known domains like awscli.amazonaws.com and sdk.cloud.google.com.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as cloud resource names and metadata, creating an attack surface where maliciously named resources could influence agent behavior.
[DYNAMIC_EXECUTION]: Utilizes commands like Invoke-AzVMRunCommand to execute arbitrary scripts on remote virtual machines as described in references/detailed-guide.md.

cloud-penetration-testing