The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands and scripts for interacting with Entra ID environments. These include reconnaissance via Azure CLI (az), directory enumeration with Microsoft Graph PowerShell, and the use of specialized tools for token theft and credential dumping.
[EXTERNAL_DOWNLOADS]: The skill references several well-known third-party security tools and PowerShell modules, including AADInternals, ROADtools, AzureHound, o365spray, TREVORspray, and MSOLSpray. These are established tools within the security community for cloud identity assessment.
[PROMPT_INJECTION]: The skill exhibits surface area for indirect prompt injection as it processes untrusted data.
Ingestion points: User-supplied inputs such as target domains, user lists (users.txt), passwords, and token values are interpolated into shell and PowerShell commands in SKILL.md.
Boundary markers: Absent. The instructions do not include delimiters or warnings to the agent regarding the handling of potentially malicious content within these inputs.
Capability inventory: The skill facilitates broad capabilities including network requests (curl), local binary execution (ROADtoken.exe), and environment modifications through PowerShell module imports.
Sanitization: Absent. There is no evidence of input validation or escaping for the user-supplied parameters before they are utilized in command execution.

attacking-entra-id