The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell commands for the agent to execute, involving offensive security tools such as impacket, netexec, ldapsearch, hashcat, and john. These commands are designed to perform credential harvesting, ticket forgery, and lateral movement.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data retrieved from an Active Directory environment, which could be manipulated by an attacker to influence the agent's subsequent high-privilege actions.
Ingestion points: Active Directory LDAP search results and SPN enumeration output in SKILL.md and references/advanced-kerberoast.md.
Boundary markers: None identified; there are no instructions to ignore potentially malicious embedded content in the ingested AD data.
Capability inventory: Extensive shell command execution (impacket suite, hashcat), file system writes (storing hashes), and network operations (Kerberos ticket requests).
Sanitization: None identified; the skill directly processes external query results to build command arguments.
[DYNAMIC_EXECUTION]: The file references/advanced-kerberoast.md contains a Python script snippet that uses subprocess.run to dynamically construct and execute shell commands. It interpolates target user variables directly into the command arguments for impacket-GetUserSPNs.
[TIME_DELAYED_CONDITIONAL]: The automation script in references/advanced-kerberoast.md implements random time delays (time.sleep(random.randint(30, 90))) to gate the execution of command-line tools. This logic is explicitly intended to evade Blue Team detection mechanisms and behavior-based monitoring.

kerberoast-attack