The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides detailed exploitation steps for multiple AWS IAM privilege escalation paths, such as 'Shadow Admin' permissions (e.g., iam:CreateAccessKey, iam:AttachUserPolicy) and instructions for launching EC2 instances with administrative roles.
[PERSISTENCE]: Includes methods for maintaining access through backdooring Lambda functions and instructions for 'Covering Tracks' which involve disabling or deleting AWS CloudTrail logs to evade detection.
[DATA_EXFILTRATION]: Describes techniques for exfiltrating sensitive data, including harvesting temporary credentials via SSRF on the EC2 metadata endpoint (IMDSv1 and v2), syncing entire S3 bucket contents to local storage, and mounting EBS snapshots to extract secrets.
[EXTERNAL_DOWNLOADS]: Fetches numerous offensive security and auditing tools from various GitHub repositories, including Pacu from Rhino Security Labs and CloudMapper from Duo Labs.
[COMMAND_EXECUTION]: Utilizes extensive shell command sequences for cloud resource enumeration, exploitation using the AWS CLI, and remote command execution on EC2 instances via AWS Systems Manager (SSM).
[INDIRECT_PROMPT_INJECTION]: Ingests and processes untrusted data from cloud service responses and metadata endpoints without sanitization or boundary markers, creating a surface for potential instruction injection via data returned by AWS APIs.
Ingestion points: AWS CLI command outputs (e.g., sts get-caller-identity, iam list-users) and HTTP responses from the EC2 metadata service.
Boundary markers: None present in the instructions to delimit external data.
Capability inventory: Extensive file-system write access, network operations, and administrative command execution capabilities across all scripts.
Sanitization: No evidence of sanitization or filtering for external content before interpolation into agent context.

aws-penetration-testing