Skip to content

Intelligence Tools

This page covers the user-facing “Intelligence” workspace features that shape chat context, reusable behavior, and output style.

Implemented behavior lets users curate connector-backed files or data, upload device files into eligible connected providers, and use Library contents as chat focus. Library focus is mutually exclusive with connector/file/folder focus.

Current limits: website attachments, sync validation cleanup, and direct chat-composer uploads are deferred.

Implemented V1 behavior lets users create, publish, archive, and select one active skill in chat. Published skill versions are immutable runtime artifacts mounted through Mastra workspace skill tooling; mutable drafts do not shape runtime behavior.

Current limits: arbitrary multi-skill stacking, cross-user skill sharing, and user-defined tool grants are out of V1 scope.

Implemented V1 behavior lets users create Markdown response templates and select or clear one template in chat. Templates shape tone and structure only; they do not expand tool access, data access, model access, focus, or chat mode policy.

Current limits: template sharing, org-wide template libraries, archived-template recovery, and rendered preview are deferred.

The current web shell exposes Chats, Connectors, Library, Response Templates, Skills, Spaces, and account controls through shared navigation and responsive layout patterns.

Current limits: focused-chat boundary hints, branded error routes, and whitelabel consumption are deferred.

  • docs/features/library/spec.md
  • docs/features/skills/spec.md
  • docs/features/skills/decisions/ADR-001-published-mastra-skill-runtime.md
  • docs/features/response-templates/spec.md
  • docs/features/web-workspace/spec.md
  • apps/web/app/pages/library
  • apps/web/app/pages/skills
  • apps/web/app/pages/response-templates
  • apps/api/src/mastra/skills