The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands for interacting with Aliyun infrastructure, including opening security group ports via AuthorizeSecurityGroup and executing arbitrary shell commands on ECS instances using the Aliyun Cloud Assistant (RunCommand) tool.
[PRIVILEGE_ESCALATION]: Instructions include detailed techniques for escalating privileges within the Aliyun Resource Access Management (RAM) service, such as attaching the AdministratorAccess policy to the current user or generating new AccessKeys for other high-privilege users.
[DATA_EXFILTRATION]: The methodology describes how to locate and download sensitive files from OSS buckets (including .env, .key, and .sql backups) and how to create database accounts to extract information from RDS instances.
[REMOTE_CODE_EXECUTION]: Documents the process for creating privileged Kubernetes pods in ACK clusters that mount the host's root filesystem, facilitating container escape and command execution on the underlying node.
[EXTERNAL_DOWNLOADS]: Recommends the installation and use of third-party exploitation frameworks from untrusted external sources, specifically the cf (Cloud Exploitation Framework) tool from github.com/teamssix/cf.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it instructs the agent to read data from external, untrusted sources such as OSS objects, database records, and application logs which could contain malicious payloads designed to manipulate the agent's behavior.
Ingestion points: Commands for reading OSS bucket contents, RDS database records, and SLS logs.
Boundary markers: None identified.
Capability inventory: Extensive capabilities including cloud resource management, command execution, and IAM policy modification.
Sanitization: No mentions of input validation or sanitization for the data retrieved from external services.
[PERSISTENCE]: Documents methods for maintaining access, such as creating new database accounts for persistent access and updating Function Compute (FC) code to include backdoor functionality.
[DYNAMIC_EXECUTION]: Includes patterns for generating and executing shell scripts on target instances via cloud management APIs.

aliyun-pentesting