The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation and the scripts/agent.py file provide functionality to execute commands on remote systems using wmiexec.py and dcomexec.py. This capability allows for remote code execution on target hosts during red team operations.\n- [PERSISTENCE]: The skill provides explicit instructions for creating WMI event subscriptions (__EventFilter, CommandLineEventConsumer) to execute arbitrary payloads on specific system triggers, such as user logons. This is a common technique for maintaining long-term access to a compromised system.\n- [COMMAND_EXECUTION]: The scripts/agent.py script employs subprocess.run() to execute external commands and tools, including tshark and native wmic commands. This enables the agent to execute shell commands on the host machine to perform audits and network analysis.\n- [CREDENTIALS_UNSAFE]: The skill facilitates credential harvesting by documenting the use of reg save to dump the SAM and SYSTEM registry hives for offline hash extraction. It also provides commands for Pass-the-Hash authentication using wmiexec.py.\n- [METADATA_POISONING]: The SKILL.md identifies the author as "mahipal", whereas the LICENSE file attributes copyright to "mukul975". This discrepancy in metadata fields can lead to confusion regarding the authorship and source of the skill.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data that could potentially contain malicious instructions.\n
Ingestion points: scripts/agent.py parses EVTX logs and PCAP files; scripts/process.py loads JSON movement logs.\n
Boundary markers: No explicit delimiters or instructions are used to distinguish data from instructions during processing.\n
Capability inventory: The agent.py script has access to shell execution via subprocess.run.\n
Sanitization: There is no evidence of sanitization for the content of the logs being processed, which could lead to an agent following instructions embedded within log entries.

performing-lateral-movement-with-wmiexec