The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides documentation on handling external data via loaders and actions, which constitutes an attack surface for indirect prompt injection if the resulting application processes untrusted data without proper boundaries.
Ingestion points: Data ingestion occurs through loader and action exports in route modules, as described in references/start/data-loading.md and references/start/actions.md.
Boundary markers: The documentation does not specify boundary markers or instructions for AI agents to separate untrusted data from instructions when processing these inputs.
Capability inventory: Applications built using these patterns have capabilities to perform server-side data access, database mutations, and external network requests via fetch as shown in references/explanation/backend-for-frontend.md.
Sanitization: The documentation includes security best practices, such as utilizing escapeHtml for output encoding and performing server-side validation in actions as documented in references/how-to/form-validation.md.

react-router