The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's installation instructions in metadata.json involve downloading content from an untrusted GitHub repository (majiayu000/claude-skill-registry) using curl.
[COMMAND_EXECUTION]: The skill requires the windows-command-line tool to execute high-privilege PowerShell commands such as Get-ADUser, Unlock-ADAccount, and Get-ADComputer. These commands interact directly with sensitive directory services.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection (Category 8).
Ingestion points: Data enters the agent context through the output of PowerShell cmdlets (Get-ADUser, Get-ADGroupMember, Get-ADOrganizationalUnit) and ADSI searchers defined in SKILL.md.
Boundary markers: None identified; the agent processes raw command output without delimiters to distinguish data from instructions.
Capability inventory: The skill uses the windows-command-line tool, allowing for arbitrary command execution on the host system.
Sanitization: No evidence of output sanitization or validation of AD object attributes (e.g., 'Description' or 'DisplayName' fields) which could be used by an attacker to inject instructions.

active-directory