vmware-vcenter-attack

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains numerous shell commands for the agent to execute against external targets. These include using curl for HTTP probing, openssl for TLS certificate metadata extraction, and ldapsearch for directory enumeration. Evidence: Found in multiple steps throughout SKILL.md, specifically for service discovery and vulnerability validation.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates an attack surface where the agent ingests data from external network responses which could contain malicious prompts. 1. Ingestion points: Responses from curl and ldapsearch commands (e.g., Step 1, Step 4, and Step 9). 2. Boundary markers: The instructions do not specify any delimiters or safety markers to isolate target data from the agent's control logic. 3. Capability inventory: Shell execution capabilities (curl, openssl, ldapsearch) and potential file access via datastore enumeration. 4. Sanitization: There is no requirement for the agent to sanitize or validate the external response data before processing.
  • [CREDENTIALS_UNSAFE]: The skill includes a list of hardcoded default credentials for VMware products to be used in authentication attempts. Evidence: Step 5 contains a table listing default passwords like vmware for legacy systems and Aria Operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 05:32 AM
Security Audit — agent-trust-hub — vmware-vcenter-attack