The Agent Skills Directory

[SAFE]: The skill provides architectural guidance and implementation templates for Microsoft Orleans. All code snippets use industry-standard security practices, such as DefaultAzureCredential for Azure services and generic placeholders for connection strings.
[INDIRECT_PROMPT_INJECTION]: The skill involves the agent processing user-provided code solutions and runtime logs, which is a potential surface for indirect prompt injection. The skill includes a 'Diagnostic Output Budget' section in SKILL.md that explicitly instructs the agent to cap diagnostic data at 8 KiB or 80 lines and filter out unnecessary logs, serving as a mitigation against context-manipulation attacks.
Ingestion points: SKILL.md (Workflow: Inspecting user solutions, interfaces, and logs).
Boundary markers: Present in SKILL.md (Diagnostic Output Budget limits).
Capability inventory: The agent performs code inspection, implementation generation, and validation.
Sanitization: Structural filtering is implemented through the diagnostic budget.
[EXTERNAL_DOWNLOADS]: The skill references official Microsoft Orleans NuGet packages and documentation links from trusted sources, including Microsoft Learn and the official .NET GitHub repositories.
Evidence: references/persistence-api.md and references/scheduling-and-services.md suggest the installation of official framework packages like Microsoft.Orleans.Journaling.

orleans