The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow in SKILL.md utilizes PowerShell commands such as New-ADUser and Add-ADGroupMember to create honeytoken accounts in Active Directory environments.
[INDIRECT_PROMPT_INJECTION]: The scripts/agent.py honeypot script ingests data from external HTTP requests, creating a potential log injection or indirect prompt injection surface.
Ingestion points: HoneypotHTTPHandler.do_GET and do_POST in scripts/agent.py read request paths, headers, and body content.
Boundary markers: None.
Capability inventory: File system access for decoy creation and network listening for the honeypot server.
Sanitization: The script does not sanitize the captured request data before logging or storing it.

performing-deception-technology-deployment