The Agent Skills Directory

[COMMAND_EXECUTION]: Modifies the system's /etc/hosts file to add DNS entries for target infrastructure, which is a sensitive system configuration change.
[COMMAND_EXECUTION]: Accesses highly sensitive system files and data stores, including the Windows WMI repository (OBJECTS.DATA) and Kerberos credential caches (.ccache files).
[COMMAND_EXECUTION]: Utilizes powerful offensive security tools such as ntlmrelayx.py, PetitPotam.py, and Mimikatz to perform authentication coercion, credential relaying, and secret harvesting.
[DATA_EXFILTRATION]: Performs large-scale harvesting of sensitive credentials from SCCM infrastructure, including Network Access Account (NAA) credentials and OSD task sequence secrets.
[CREDENTIALS_UNSAFE]: Includes hardcoded placeholder credentials (e.g., 'Password123', 'Password123!') in command-line examples for authentication and machine account creation.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection due to the ingestion and processing of untrusted data from network-accessible SCCM management points (XML policy files and HTTP responses) without boundary markers or sanitization.
Ingestion points: Fetches XML policy data via sccmwtf.py and retrieves site information from Management Point HTTP endpoints.
Boundary markers: Lacks delimiters or warnings to prevent the agent from interpreting embedded data as instructions.
Capability inventory: Has extensive access to shell execution, file system operations, and network tools.
Sanitization: No evidence of data validation, escaping, or filtering for ingested content.

sccm-exploitation