The Agent Skills Directory

[Security-Centric Guidance]: The skill actively encourages security best practices, such as using Microsoft Entra ID for passwordless authentication and managed identities instead of hardcoded credentials. It explicitly advises against storing secrets in repository configurations.
[Prompt Processing Surface]: The skill identifies user intents (e.g., "build an app on Azure SQL") to route tasks. Like any skill processing natural language prompts, this represents a surface for indirect prompt injection.
Ingestion points: User prompts described in the skill metadata (SKILL.md).
Boundary markers: No explicit boundary delimiters or "ignore" instructions are present for user input interpolation.
Capability inventory: The skill uses az CLI for infrastructure management, sqlcmd for database interaction, and package managers like npm and dotnet for environment setup.
Sanitization: The skill refers to a separate prevent-sql-injection skill for sanitizing query parameters.
[External Dependency Management]: The instructions involve installing packages from official registries (e.g., Prisma from NPM, Microsoft.Data.SqlClient from NuGet) and Azure CLI extensions like serviceconnector-passwordless. These resources originate from well-known and trusted providers.
[Infrastructure Verification]: The skill uses read-only CLI commands to verify firewall rules and administrative settings, helping users avoid common configuration errors that could lead to connectivity issues.

build-app-on-azure-sql