The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes patterns for ingesting and processing untrusted data from external users via API Gateway (headers, body, path parameters) and pagination cursors. While this is an inherent part of API functionality, it represents a surface area for indirect injection.
Ingestion points: API request components are processed in references/middleware-pattern.md, and base64-encoded cursors are handled in references/utilities-pattern.md.
Boundary markers: The pattern explicitly recommends using Zod-based validation middleware (validateBody, validateHeaders, validatePathParameters) to enforce input schemas.
Capability inventory: Lambda functions are granted scoped access to services like Amazon S3 and SES (via createSendEmailPolicy in `references/node-lambda-pattern.md"), and they interact with the persistence layer through structured repositories.
Sanitization: The skill prioritizes Zod for input validation and provides a RestResult utility for consistent error handling and response shaping.
[DYNAMIC_EXECUTION]: The pagination logic in references/utilities-pattern.md utilizes JSON.parse to decode base64-encoded strings provided by clients as cursors. While a standard implementation for continuation tokens, the skill correctly positions this within the repository layer to isolate data parsing from business logic.

cdk-rest-api