The Agent Skills Directory

[SAFE]: The skill utilizes official documentation and community-standard libraries for its functionality. All external URLs point to established and reputable domains such as nitro.build, h3.dev, and unjs.io.
[EXTERNAL_DOWNLOADS]: The instructions reference fetching configuration and guidelines from official GitHub repositories and project websites. These sources are recognized as established entities within the web development ecosystem.
[COMMAND_EXECUTION]: The skill provides instructions for standard development operations using the nitro CLI, vite, and package managers. These actions are performed through common commands like bunx nitro build which are typical for the framework's workflow.
[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for building servers that process untrusted data, creating a potential attack surface.
Ingestion points: Processes URL parameters, headers, and request bodies in routes/ (routing.md) and server entry files (server-entry.md), as well as WebSocket messages (tasks-openapi-websocket.md).
Boundary markers: No explicit delimiters for untrusted input are mandated in the instruction set; it follows standard web request handling conventions.
Capability inventory: Handlers can perform network requests via fetch (cache-storage-database.md), write to build directories (SKILL.md), and execute internal tasks (tasks-openapi-websocket.md).
Sanitization: The documentation recommends implementing authentication for production endpoints, specifically for OpenAPI UIs and HTTP-triggered tasks (tasks-openapi-websocket.md).

nitro