cloud-attack-methods
Warn
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONPERSISTENCEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documents numerous commands for unauthorized resource enumeration and exploitation, such as
aws s3 ls --no-sign-request,kubectl auth can-i --list, and brute-forcing MinIO console logins via API POST requests. - [PRIVILEGE_ESCALATION]: Extensive details are provided for escalating privileges in cloud environments, including 18+ paths for AWS IAM (e.g.,
PassRole), Azure Managed Identity token theft via IMDS, and GCP service account token creation. - [PERSISTENCE]: The skill describes methods for establishing stealthy persistence in Entra ID (Azure AD), specifically through the addition of credentials to service principals and the abuse of dynamic groups.
- [DATA_EXFILTRATION]: Techniques for extracting sensitive data are outlined, including reading environment variables from Lambda functions, harvesting tokens from metadata APIs (169.254.169.254), and using path traversal (CVE-2023-28434) to read
/etc/passwd. - [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where an agent might be influenced to perform malicious actions by following the detailed tactical instructions provided.
- Ingestion points: The
SKILL.mdfile contains a high concentration of offensive payloads and techniques. - Boundary markers: The instructions lack any boundary markers or warnings to distinguish documentation from executable intent.
- Capability inventory: The skill assumes access to high-privilege tools such as
kubectl,aws-cli,gcloud, and specialized pentesting frameworks. - Sanitization: There is no sanitization or validation logic for the external attack paths and tools described.
Audit Metadata