The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents an orchestration framework where agents (actors) ingest and process data from external queues, connections, and HTTP requests.
Ingestion points: Multiple examples, such as examples/docs/actors-queues/multiple-queues/index.ts, demonstrate actors retrieving and processing data payloads from internal message queues.
Boundary markers: As the skill provides a development kit for third-party implementation, the examples focus on logic rather than specific prompt delimiters, which is standard for architectural documentation.
Capability inventory: Actors in the examples are granted capabilities for network operations (via fetch), persistent storage access (via c.kv), and event broadcasting to connected clients.
Sanitization: The skill promotes secure development by providing dedicated examples for input validation, such as using the zod library to define and enforce strict schemas for incoming data in examples/docs/actors-input/input-validation.ts.- [SAFE]: All external URLs and package references are associated with the verified vendor (rivet.dev) or well-known, trusted open-source libraries. The code is transparent and follows established patterns for agent orchestration frameworks.

rivet-actors