The Agent Skills Directory

[DYNAMIC_EXECUTION]: The ConfigureConfigDiscovery method in discovery-providers.md uses string interpolation to assemble a HOCON (Typesafe Config) string from variables such as options.ServiceName and endpoints. This generated configuration string is then passed to builder.AddHocon(hocon, ...) for processing by the ActorSystem. Assembling executable configuration strings from variables without explicit sanitization creates a risk of configuration injection if the input values originate from untrusted sources.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates service discovery by ingesting data from external sources, specifically the Kubernetes API and Azure Table Storage. This configuration creates a vulnerability surface for indirect injection.
Ingestion points: Kubernetes pod metadata (labels/ports) and Azure Table Storage rows are read by the discovery providers defined in discovery-providers.md.
Boundary markers: The provided configuration logic lacks explicit delimiters or validation to distinguish between legitimate discovery data and potentially malicious embedded instructions.
Capability inventory: The skill enables network port binding (port 8558 for management), cluster membership management, and ActorSystem initialization which can involve remote messaging and state changes across the cluster.
Sanitization: There is no evidence of sanitization or escaping for the discovery data before it is processed by the Akka.Management bootstrap mechanism.
[PRIVILEGE_ESCALATION]: The discovery-providers.md file includes a Kubernetes Role definition that grants get, list, and watch permissions on pods to the application's ServiceAccount. While these permissions are required for the Akka.Discovery.KubernetesApi provider to function, granting broad read access to pod metadata in a namespace increases the privilege level and potential blast radius of the application.

akka-net-management