The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and analyze untrusted application code and configuration files (e.g., source manifests and environment configurations) from various cloud platforms. This capability creates a potential surface for indirect prompt injection, as instructions embedded within the analyzed data could attempt to influence the agent's migration logic. To mitigate this risk, the skill mandates a sequential workflow requiring a manual assessment report before any code migration is performed, and it includes rules for auditing hardcoded hostnames to ensure sanitization of the resulting output.
Secure Secret Management Pattern: A core function of the skill involves the migration of sensitive credentials from source providers to Azure Key Vault. To prevent accidental exposure of these secrets in shell history or process logs, the skill provides high-quality guidance for secure handling, such as using temporary files with restricted permissions and implementing automatic cleanup via secure deletion methods. This avoids the insecure practice of passing secrets as direct command-line arguments.
Identity-First Authentication: The skill promotes a security-first approach by prioritizing Microsoft Entra ID and Managed Identities over traditional, long-lived credentials such as API keys or connection strings. By providing code patterns that utilize the DefaultAzureCredential, the skill helps reduce the likelihood of credential theft or exposure during and after the migration.
Controlled Environment Modification: While the skill provides extensive automation for resource deployment and configuration, it implements global safety rules requiring explicit user confirmation before any destructive actions or production-level changes are made, maintaining human oversight of the process.

azure-cloud-migrate