The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The scripts/manage_ad_users.ps1 script contains a hardcoded plaintext password (TempPassword123!) within the New-ADUserAdvanced function. This password is automatically applied to new Active Directory accounts, creating a significant security risk for new users.
[DATA_EXFILTRATION]: The scripts/configure_dns.ps1 script defaults to using Google Public DNS (8.8.8.8) for name resolution queries in the Get-DNSQuery function. This configuration allows the skill to resolve names using a well-known external service.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from users to perform critical administrative operations, which creates a surface for potential exploitation via indirect prompt injection.
Ingestion points: User-provided inputs for usernames, emails, and Group Policy object names across the administrative scripts.
Boundary markers: None; the skill does not use delimiters to isolate external data from its administrative commands.
Capability inventory: The skill is capable of modifying Active Directory objects, creating DNS records, and applying system-wide security policies.
Sanitization: Validation is present for specific string formats such as email addresses, but the skill lacks comprehensive sanitization to protect against malicious input influencing administrative workflows.
[DYNAMIC_EXECUTION]: The scripts/setup_gpo.ps1 script dynamically assembles security configuration files (.inf) by interpolating parameters into a template and applies these configurations to the system using the secedit utility. This approach to runtime configuration generation is used for managing security policies.

windows-infra-admin