The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines a set of behaviors for the agent to follow when analyzing or generating backend code. It relies on the agent processing user-provided source code (e.g., routes, controllers, and services).
Ingestion points: User-supplied code files and terminal input during code reviews or refactoring tasks.
Boundary markers: The skill does not explicitly define delimiters to separate user-provided code from the agent's instructions.
Capability inventory: The agent operating under these guidelines typically possesses file system access and terminal execution capabilities.
Sanitization: Input is treated as source code for analysis; no specific sanitization or validation is applied to the raw input before processing by the LLM.
[CREDENTIALS_UNSAFE]: The documentation includes examples of configuration files (e.g., config.ini, .env) that contain placeholder keys and secrets such as password1, your-jwt-secret, and dev-secret. These are clearly identifiable as illustrative examples and do not represent actual exposed credentials.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known libraries and services such as Prisma, Zod, and Sentry. These are standard industry tools for the stated purpose of backend development and monitoring.

backend-dev-guidelines