The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a variety of shell commands, including curl, openssl, and ldapsearch, to perform reconnaissance and data gathering against network targets.
[REMOTE_CODE_EXECUTION]: Step 4 provides a functional exploit payload for a Server-Side Template Injection (SSTI) vulnerability in VMware Workspace ONE Access (CVE-2022-22954), allowing the execution of arbitrary commands.
[CREDENTIALS_UNSAFE]: Step 5 lists several hardcoded default credentials for VMware components, such as root:vmware, which are commonly targeted in automated attacks.
[DATA_EXFILTRATION]: Step 8 details how to locate and download sensitive datastore files, including VMDKs and snapshots, which may lead to significant data loss.
[INDIRECT_PROMPT_INJECTION]: (1) Ingestion points: Data is received from external target endpoints in Steps 1, 4, and 6 through 10. (2) Boundary markers: There are no instructions to the agent to treat external server responses as untrusted content. (3) Capability inventory: The skill utilizes network commands and local file system access based on external inputs. (4) Sanitization: The skill does not implement any validation or sanitization of the data retrieved from targets.

vmware-vcenter-attack