The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides automated scripts that modify the system's security posture by changing the PowerShell Execution Policy to permit script execution.\n
Evidence: The Bootstrap-ScoopEnvironment.ps1 script in SKILL.md executes Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser -Force.\n- [PRIVILEGE_ESCALATION]: The skill instructs the user or agent to disable Windows Defender antivirus scanning for the Scoop installation directory.\n
Evidence: In gemini.md, the skill recommends running Add-MpPreference -ExclusionPath "$env:USERPROFILE\scoop" to avoid performance issues, which creates a security blind spot for any binaries managed by the package manager.\n- [REMOTE_CODE_EXECUTION]: The skill employs a piped command pattern to download and execute the Scoop installer from a remote source.\n
Evidence: SKILL.md uses Invoke-RestMethod -Uri "https://get.scoop.sh" | Invoke-Expression. While targeting the official installer of a well-known service, this method executes remote code directly in memory without prior verification.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external, third-party package manifests.\n
Ingestion points: The skill reads JSON manifests and Git bucket metadata from remote repositories during scoop bucket add and scoop install operations.\n
Boundary markers: None identified in the instructional content.\n
Capability inventory: The skill has extensive system capabilities including shell execution, security policy modification (ExecutionPolicy, MpPreference), and file system access across the user profile.\n
Sanitization: No sanitization or integrity validation of the external manifest content is described.

scoop