The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data through routing composables and server function validators, creating a potential surface for indirect prompt injection.
Ingestion points: External data enters the agent context via useParams, useSearch, and useLoaderData in SKILL.md (e.g., in src/routes/index.tsx).
Boundary markers: The skill demonstrates the use of a .validator() in createServerFn to enforce data types, but lacks explicit instructions or markers to prevent the agent from interpreting embedded instructions within this data.
Capability inventory: The skill provides capabilities for database operations (e.g., db.posts.update in the useServerFn example) and server-side code execution via createServerFn across all component and route files.
Sanitization: There is no explicit sanitization or filtering logic shown to prevent untrusted external data from influencing agent behavior during runtime processing.

tanstack-vue-start