The Agent Skills Directory

[SAFE]: The skill explicitly enforces a read-only policy for all catalog operations, forbidding any mutating actions such as creation or deletion of resources. This adherence to least privilege prevents unauthorized infrastructure changes.
[SAFE]: The skill mandates high data integrity by requiring direct quotes from tool outputs and reporting 'none found' rather than fabricating information, which prevents model hallucinations.
[SAFE]: While the skill ingests metadata from external repositories (Hub and public templates), it lacks exploitable capabilities such as file system writes, subprocess execution, or arbitrary network access, rendering the surface for indirect prompt injection non-exploitable.

runpod-discovery