The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py utilizes the subprocess module to execute local forensic tools including Volatility 3, Sysmon (wevtutil), and PowerShell. These calls are implemented using list-based arguments and static command strings, which prevents common shell injection vulnerabilities by ensuring that user-provided parameters (such as PIDs or file paths) are not interpreted as commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data from memory dumps and system event logs (EVTX). Maliciously crafted system logs or process names could contain strings designed to influence the AI agent's behavior when it reviews the generated analysis report.
Ingestion points: Data is read from EVTX files in scripts/agent.py (within scan_sysmon_injection_events) and from Volatility command output.
Boundary markers: The script does not utilize specific delimiters or warnings to separate untrusted log content from the rest of the agent's context.
Capability inventory: The script can perform file reads and execute subprocesses for forensic querying.
Sanitization: There is no evidence of sanitization or filtering for the strings extracted from system logs before they are formatted into the final JSON report.

detecting-process-injection-techniques