The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the installation of official Inngest middleware packages such as @inngest/middleware-encryption and @inngest/middleware-sentry. It also documents standard dependencies including openai, redis, stripe, and @prisma/client. These are established industry libraries and official platform extensions.\n- [DATA_EXFILTRATION]: A code template is provided for custom error tracking (createErrorTrackingMiddleware) that uses the fetch API to send execution context and event data to a user-defined endpoint. This is presented as a developer utility for observability and relies on user-supplied configuration for the destination.\n- [CREDENTIALS_UNSAFE]: All code examples correctly demonstrate the use of environment variables (e.g., process.env.ENCRYPTION_KEY, process.env.STRIPE_SECRET_KEY) to handle sensitive API keys and encryption secrets, avoiding hardcoded credentials.\n- [PROMPT_INJECTION]: The skill describes a middleware system that processes untrusted data payloads, which is an inherent attack surface for indirect prompt injection.\n
Ingestion points: ctx.event.data (documented in references/built-in-middleware.md and references/dependency-injection.md).\n
Boundary markers: None are present in the provided implementation templates.\n
Capability inventory: Network operations via fetch, database interactions via Prisma or SQLAlchemy, and external service calls via OpenAI and Stripe clients.\n
Sanitization: The provided examples focus on functionality and do not include explicit sanitization logic for the incoming event data.

inngest-middleware