The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data when instructed to "triage BloodHound output," which represents a potential attack surface.
Ingestion points: The agent is expected to ingest and analyze JSON or CSV output from Active Directory graph analysis tools like BloodHound (SharpHound, Rusthound).
Boundary markers: The instructions do not define specific delimiters or "ignore previous instructions" markers for the ingested data.
Capability inventory: The skill is scoped to analysis, ranking attack paths, and generating reports; it does not include tools for automated shell execution or file writing based on the data.
Sanitization: There are no explicit instructions for the agent to sanitize or validate the content of the ingested data files.
[EXTERNAL_DOWNLOADS]: The skill references several external repositories and tools as resources for the user.
Evidence: Links to the official Microsoft repository for New-KrbtgtKeys.ps1 and the GhostPack repository for PSPKIAudit are provided as documentation.
Note: These references target well-known services and organizations and are intended for manual use in a lab environment rather than automated execution.

ad-attacks