Cloud Penetration Testing

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEDATA_EXFILTRATIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches and installs official cloud command-line interfaces from trusted Google Cloud and Amazon Web Services (AWS) domains to set up the testing environment.\n- [REMOTE_CODE_EXECUTION]: Executes installation scripts for the Google Cloud SDK and AWS CLI downloaded from official provider domains (google.com and amazonaws.com).\n- [COMMAND_EXECUTION]: Utilizes industry-standard cloud CLI tools (az, aws, gcloud) and PowerShell modules to perform resource enumeration, security audits, and remote command execution on virtual machines.\n- [PERSISTENCE]: Outlines methods for establishing persistent access within target cloud environments, such as creating backdoor service principals in Azure AD and generating additional IAM access keys in AWS for testing purposes.\n- [DATA_EXFILTRATION]: Details techniques for retrieving sensitive information from cloud targets, including the extraction of Azure Key Vault secrets, AWS Lambda environment variables, and local cloud CLI configuration files containing credentials.\n- [PRIVILEGE_ESCALATION]: Includes strategies for discovering and exploiting misconfigured IAM roles and permissions to gain elevated access within a cloud tenant or project.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input data (e.g., lists of IP addresses and usernames) from local files like ips.txt and users.txt without explicit boundary markers or sanitization logic, representing a vulnerability surface for potential instruction injection during automated workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 03:31 PM