Example Catalog

Use these names as routing examples, not as promises of a new platform capability. Each starts from the smallest current template and follows the draft-only lifecycle in validation-and-lifecycle.md.

Golden example	Route to	Core boundary
worked-walkthrough	python-api	End-to-end application of every skill step to one request; also the minimal single-Converse-call reference.
live-operations-dashboard	multi-source-dashboard + live SQL datasets	Bounded joins/metrics and visible TTL freshness.
viewer-scoped-customer-portal	viewer-scoped-dashboard	Signed-in viewer; currently only customers.email row filtering.
internal-action-tool	internal-tool	Private tool; use declared server routes and guarded workspace state only when needed.
private-api-first-app	python-api	Private /_artifact/api/* routes and server-side integrations.
customer-scoped-map	Platform-service-backed UI	MapLibre browser surface over a tenant-authenticated, customer-account Amazon Location Maps and Places broker; no browser credentials or implicit location writes.
FastMCP server	fastmcp-server	Live-only exact MCP ingress, platform-owned credential, and customer-managed tool review for chat/coding.
bounded-bedrock-assistant	python-api	Private route, one bounded Bedrock Converse call, no agentic extras.
knowledge-base-backed-app	react-dashboard + an exact published Knowledge Base grant	Normal artifact reads one approved live release through relative /_artifact/knowledge/* broker routes; no token, provider ID, or direct public endpoint.
signed-webhook-boundary	Platform service boundary	Current artifact routing cannot expose provider-verifiable, durable machine ingress.
cache-warmed-multi-source-dashboard	multi-source-dashboard + live dataset	Explicit recurring refresh_live_cache, Redis preflight, live fallback, and visible generatedAt.
precomputed-s3-report	artifact-draft or react-dashboard + precomputed dataset	Platform-managed private object storage through declared routes; never expose object URLs.
materialized-sql-data-product	multi-source-dashboard after upstream work	Upstream materialized table/view is separately owned and cataloged; app performs bounded read-only access.
customer-api-source-pipeline	Data Sources build + platform.connector.rest.v1	The operational connector, credentials, checkpoints and tables are first-class Data Source objects; with the workspace on, scheduling it creates its data pipeline artifact.
data-pipeline-artifact	Built-in data pipeline artifact (no template)	One artifact per connector instance: tables and named schedules (including PestRoutes hard-delete schedules) in a pipelineManifest; the coder saves drafts, a person publishes.

Template choice may change if the visual surface changes, but the stated security and data boundary must not. In particular, do not use the aggregate template for multi-source math, use a precompute schedule merely as a cache warmer, or turn a webhook into a viewer-authenticated route.