The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes robust defensive instructions to ignore attempts to override agent behavior. It explicitly warns the agent to treat all user-provided content (task descriptions, file names, environment variables) as untrusted data and to halt rather than comply if instruction-like patterns are detected in user input.
[COMMAND_EXECUTION]: Shell command safety is a core component of the skill's design. It restricts execution to an approved set of tools (swa, az, npm, git) and provides explicit instructions to never construct shell arguments by concatenating unvalidated user-supplied strings.
[CREDENTIALS_UNSAFE]: The skill promotes secure credential handling across all resources. It provides examples using GitHub Secrets (${{ secrets.AZURE_STATIC_WEB_APPS_API_TOKEN }}) for CI/CD and explicitly warns users to add local.settings.json to .gitignore to prevent accidental credential leakage. It also recommends using Azure Key Vault for production secrets.
[INDIRECT_PROMPT_INJECTION]: The skill identifies clear attack surfaces where untrusted data could influence behavior (e.g., writing staticwebapp.config.json). It provides a mandatory sanitization table with specific regex patterns and validation rules for route patterns, role names, and environment variables to mitigate injection risks during file writing.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known tools and packages such as the official Azure Static Web Apps CLI (@azure/static-web-apps-cli) and official GitHub Actions from the Azure/ organization. These are recognized as trusted industry sources.

azure-swa