The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute high-privilege system commands for recovery purposes.
Evidence: Use of wbadmin for Active Directory restoration, Set-ADAccountPassword for credential resets, and Start-VBRInstantRecovery for virtual machine restoration in SKILL.md.
[CREDENTIALS_UNSAFE]: The skill contains hardcoded strings used as passwords in script examples.
Evidence: SKILL.md includes strings like "NewKrbtgt2026!Complex#1" and "NewKrbtgt2026!Complex#2" within PowerShell commands meant for resetting the krbtgt account.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface through the processing of untrusted external data during the forensic scan phase.
Ingestion points: External backup data is mounted and scanned for artifacts at /mnt/backup_verify in SKILL.md (Step 3).
Boundary markers: Absent; there are no specific instructions or delimiters to isolate the output of scanning tools (clamscan, find) when processed by the agent.
Capability inventory: The skill has extensive capabilities including shell command execution (bash, powershell), file system modifications, and network operations via backup tools.
Sanitization: The skill uses the noexec mount flag in Step 3, which prevents binary execution from the backup but does not mitigate prompt injection via log output parsing.

recovering-from-ransomware-attack