feat(build): add required owner_team field to template entries

.claude/skills/template/SKILL.md

@@ -5,7 +5,7 @@ description: Create, maintain, format, and publish Anyscale console templates. U
 
 # Template skill
 
-A console template = a `BUILD.yaml` entry + `templates/<name>/` content + `configs/<name>/` compute configs + a `tests/<name>/` test. Every template under `templates/` is tested (the validator enforces it); `archive/` holds **test-exempt** content — retired/past-event templates, plus fast untested iteration that can still be published — via `workflows/archive-template.md`.
+A console template = a `BUILD.yaml` entry + `templates/<name>/` content + `configs/<name>/` compute configs + a `tests/<name>/` test. Every entry carries a required `owner_team` (`ray-serve` | `ray-data` | `llm` | `ray-train` | `general`), deduced from the template's content — the rule + tie-break live in `schemas/build-yaml-schema.yaml`. Every template under `templates/` is tested (the validator enforces it); `archive/` holds **test-exempt** content — retired/past-event templates, plus fast untested iteration that can still be published — via `workflows/archive-template.md`.
 
 **How to use:** create/update are **interactive** — interview the user, explain at the point of need, and ask for any missing input. **ray-bump is non-interactive** (the `template-updater` Cursor cloud agent; signals: `.cursor/` setup, a `cursor/...` branch, an automation trigger): never prompt, use defaults; on missing input or preflight failure, post to the PR and stop. Track multi-step runs with your task tool.
 

.claude/skills/template/references/conventions.md

@@ -23,7 +23,7 @@ templates/
 
 Run the `check-build-yaml` hook first (see **Validate locally**) — it authoritatively covers BUILD.yaml schema, referenced paths, naming, and the compute-config schema check. Fix what it reports, then verify:
 
-- **BUILD.yaml entry** matches `../schemas/build-yaml-schema.yaml`. Under `cluster_env:`, use either `cluster_env.image_uri` OR `cluster_env.byod`, never both. Image taxonomy: SKILL.md "Image URI cases".
+- **BUILD.yaml entry** matches `../schemas/build-yaml-schema.yaml`. `owner_team` is required (`ray-serve` | `ray-data` | `llm` | `ray-train` | `general`) — deduce per that schema file's rule. Under `cluster_env:`, use either `cluster_env.image_uri` OR `cluster_env.byod`, never both. Image taxonomy: SKILL.md "Image URI cases".
 - **Compute configs** present at `configs/<name>/aws.yaml` and `configs/<name>/gce.yaml`. Schema: `../schemas/compute-config-schema.yaml`.
 - **Test** present at `tests/<name>/tests.sh`.
 - **Dependencies pinned.** Declare template deps in `templates/<name>/requirements.txt`, the BYOD `Dockerfile` (`RUN pip install "pkg==x.y.z"`), or notebook `!pip install` — with exact versions. (The repo-root `dependencies/` directory is Ray base-image constraint management, not per-template deps.)

.claude/skills/template/schemas/build-yaml-schema.yaml

@@ -8,6 +8,8 @@
 # --- Full schema with all fields ---
 
 - name: <template-name>          # REQUIRED. Unique identifier. Appears in console URL.
+  owner_team: <team>             # REQUIRED. One of: ray-serve | ray-data | llm | ray-train | general.
+                                 #   Deduced from content — see "owner_team deduction" below.
   dir: templates/<dir-name>      # REQUIRED. Path to template content (relative to repo root).
 
   # REQUIRED. Exactly one of: image_uri, or byod.
@@ -59,6 +61,30 @@
 #
 # See: https://docs.anyscale.com/reference/base-images
 
+# --- owner_team deduction (content → single value) ---
+#
+# The owning team, deduced from what the template is fundamentally ABOUT (its
+# center of gravity), not merely what it imports. Pick exactly one:
+#
+#   ray-serve  - non-LLM online model serving: @serve.deployment, model
+#                composition / multiplexing, inference APIs.
+#   ray-data   - non-LLM batch/streaming data pipelines & offline batch
+#                inference: map_batches, embeddings-at-scale, ETL.
+#   llm        - fundamentally LLM: vLLM, ray.serve.llm, ray.data.llm, LLM
+#                fine-tuning, RAG, LLM agents.
+#   ray-train  - distributed training / fine-tuning (TorchTrainer, FSDP,
+#                DeepSpeed, tensor-parallel) & Ray Tune HPO, non-LLM-centric.
+#   general    - platform / Ray Core / onboarding with no single-library owner:
+#                intros (jobs, workspaces, services, clusters), multi-library
+#                surveys, pure Ray Core task/actor parallelism.
+#
+# Multi-library tie-break: llm wins when the template is fundamentally an LLM
+# template (even if it also uses ray.data for batch or ray.train to fine-tune an
+# LLM). Otherwise pick the primary purpose — a Data→Train pipeline whose point is
+# training → ray-train; a batch-inference/embeddings pipeline → ray-data; a
+# serving demo → ray-serve. Teaching the platform/Core rather than a library →
+# general.
+
 # --- Cross-field rules (enforced by ci/validate_build_yaml.py) ---
 #
 # - tests_path basename must equal name → tests/<name>/

.claude/skills/template/workflows/create-template.md

@@ -25,7 +25,7 @@ Move the content into `templates/<name>/`. For a notebook template, the main not
 
 ## 3. BUILD.yaml entry
 
-Append a list item per `../schemas/build-yaml-schema.yaml`. Set the image for the chosen case: `cluster_env.image_uri` (anyscale base) or `cluster_env.byod.{docker_image,ray_version}` (custom or third-party). For custom GCP, publish the image first (`.claude/skills/template/scripts/push-custom-image-to-gcp.sh <dockerfile-dir> <name> <ray-version>`) and use the printed URI. The entry also wires `compute_config` (step 4) and the `test` block (`command: bash tests.sh`, `tests_path: tests/<name>/`, and `timeout_in_sec` set a bit above the test's measured runtime — target < 30 min).
+Append a list item per `../schemas/build-yaml-schema.yaml`. Set `owner_team` (required — `ray-serve` | `ray-data` | `llm` | `ray-train` | `general`, deduced from the template's center of gravity; the schema file has the rule + tie-break). Set the image for the chosen case: `cluster_env.image_uri` (anyscale base) or `cluster_env.byod.{docker_image,ray_version}` (custom or third-party). For custom GCP, publish the image first (`.claude/skills/template/scripts/push-custom-image-to-gcp.sh <dockerfile-dir> <name> <ray-version>`) and use the printed URI. The entry also wires `compute_config` (step 4) and the `test` block (`command: bash tests.sh`, `tests_path: tests/<name>/`, and `timeout_in_sec` set a bit above the test's measured runtime — target < 30 min).
 
 ## 4. Compute configs