The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documentation describes a server framework that processes untrusted data from HTTP requests, which is an inherent attack surface for indirect injection.
Ingestion points: Handlers in routes/ and api/ read user-supplied data from event.req (e.g., JSON, text, form data) as documented in core-routing.md.
Boundary markers: The documentation provides explicit security guidance, such as using db.sql tagged templates for database queries and recommending authentication before running tasks.
Capability inventory: The framework supports file system access via useStorage, database operations via useDatabase, and network operations via fetch or proxying.
Sanitization: The skill recommends security best practices like using parameterized SQL queries and manual validation for task payloads to mitigate injection risks.
[EXTERNAL_DOWNLOADS]: The skill references standard development dependencies and tools within the Nitro, Nuxt, and UnJS ecosystems.
Evidence: Mentions integrations with official libraries such as h3, unstorage, db0, crossws, and scalar for API documentation.
[COMMAND_EXECUTION]: The documentation includes examples of CLI commands for development workflows.
Evidence: features-tasks.md documents standard CLI usage for listing and running tasks via nitro task list and nitro task run.

nitro