The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py executes built-in Windows commands wevtutil and powershell using subprocess.run with list-based arguments. These commands are used to collect security telemetry (Sysmon Event IDs 19, 20, 21) and enumerate WMI namespaces for active persistence mechanisms. This is the intended and necessary behavior for the skill's threat-hunting purpose.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection as it processes untrusted system event logs and WMI subscription metadata which could contain attacker-controlled strings.
Ingestion points: Untrusted event data is ingested via wevtutil and powershell outputs in scripts/agent.py.
Boundary markers: The script uses standard structured data parsers including xml.etree.ElementTree for Sysmon logs and json.loads for PowerShell output to process data.
Capability inventory: The skill performs local command execution and file writing to generate reports but does not have network exfiltration capabilities.
Sanitization: The script extracts specific telemetry fields for analysis and report generation, but does not execute the content of the data it processes as code.

detecting-wmi-persistence