ReadonlyDescriptionHuman-readable description of what this provider covers.
ReadonlyDomainOpt-in marker read by ClassFactory (see ClassResolutionResult): this class CANNOT
function standalone — every member below is abstract, so a base instance is a method-less
stub. TypeScript's abstract is erased at runtime, so the factory cannot detect that on its
own; without this marker an unresolvable ProviderClassName would silently be handed back
as a live "provider" and blow up later at the first method call.
With the marker set, CreateInstance throws and TryCreateInstance returns
{Resolved: false, Instance: null} — both of which PermissionEngine handles by skipping
the domain.
ReadonlySupportedWhat actions this provider can evaluate.
ReadonlySupportedWhat grantee types this provider supports.
ReadonlySupportsWhether this provider supports explicit Deny records.
ProtectedboolsMap a { Read?: boolean, Update?: boolean, … } descriptor to a canonical
PermissionAction[]. Replaces the 9 hand-rolled copies that each walked
their row's CRUD-ish booleans and pushed action names.
Output order matches the canonical order declared in PermissionAction.
Entries that are null/undefined/false are skipped.
ProtectedbuildBuild a NormalizedPermission with DomainName and Effect
pre-filled from the provider. Centralizes the ~10-field literal that
used to be constructed in every GetUserResources / GetEffectivePermissions /
GetResourcePermissions implementation.
Effect defaults to 'Allow' since the vast majority of providers don't
support Deny. Providers that emit Deny records (EntityPermissionProvider)
pass { effect: 'Deny' } explicitly.
ProtectedbulkGiven a list of IDs, fetch {ID, <nameField>} from entityName and return
a Map<ID, name>. Used by providers whose domain views don't denormalize
a resource name onto the permission row (AI Agents, Artifacts, Collections).
Returns an empty Map when ids is empty or the RunView fails — callers should
treat a missing key as "name unknown."
OptionalnameField: stringCheck if a user has a specific permission on a specific resource.
The user whose permissions are being checked (roles come from user.UserRoles).
The resource type within this domain (e.g., entity name, resource type name).
The specific resource ID; null for domain-wide checks.
The action being requested.
ProtectedfetchStandard RunView wrapper that logs failures with <ProviderClass>.<methodName>:
prefix and returns the row list (or [] on failure). Replaces the boilerplate
that each provider previously carried around its RunView calls.
Get all effective permissions a user has on a specific resource. Returns an empty array when the user has no access.
Default implementation returns [] — role-only providers don't support user-granted
sharing. User-granted-sharing providers (Dashboards, Artifacts, Collections, Resource
Permissions, Access Control Rules) should override this to query their source table.
Optional_provider: IMetadataProviderDefault implementation returns [] — role-only providers have no "shared with me"
concept. User-grantee providers should override this to return only permissions where
grantee is the direct grantee AND someone else is the grantor/owner.
Optional_provider: IMetadataProviderGet all permissions granted on a specific resource across every grantee. Powers the Sharing Center's "Resource Access Report" view.
Default implementation returns [] — providers should override to advertise
their supported resource types (see IPermissionProvider.GetResourceTypes).
Get all resources within this domain that the user has access to. Powers the Sharing Center's "User Access Report" view.
OptionalresourceType: stringOptional filter to one resource type within the domain.
Wraps
MJ: AI Skill Permissionsbehind the unified PermissionProviderBase contract — the closed-by-default, explicit-grants view that powers the Sharing Center and audit.This is the skill sibling of
AIAgentPermissionProvider. AISkillPermission rows grant either a User (via UserID) or a Role (via RoleID) — never both; the server entity enforces this. Action mapping:ReadExecute(request/activate the skill at agent runtime)UpdateDeleteUser+Role grants for the same user are OR-aggregated (any grant of an action suffices).
resourceTypeis"AI Skills".resourceIdis the skill ID. NOTE: this provider reports only explicit grants; the open-by-default runtime behavior (no rows → View+Run for everyone) lives inAISkillPermissionHelper— see guides/UNIFIED_PERMISSIONS_GUIDE.md for why the two paths differ.