Drop internal vocabulary, and stop shipping dev files to users

Six comments attributed decisions to "owner decision 2026-07-22", and one test
cited a path inside the private monorepo (PromptFormatDropdown.tsx). Both are
meaningless to anyone reading this repo: an outside contributor has no idea who
the owner is, and cannot open that file. Each comment already gave its reason;
only the attribution goes.

Adds .comfyignore so tests/, scripts/ and .github/ stay in the repo but stay
out of the archive published to the Registry. Installing a pack copies that
archive into custom_nodes/ and leaves it there, and none of those directories
is imported at runtime — __init__.py loads styleref_nodes/* and server_routes
only. workflows/ is deliberately still shipped; the templates are the product.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

.comfyignore

@@ -0,0 +1,24 @@
+# Excluded from the archive published to the Comfy Registry — NOT from the repo.
+#
+# Installing this pack copies the published archive into a user's custom_nodes/
+# directory, where it sits for as long as they have the pack. Nothing below is
+# imported at runtime (__init__.py loads styleref_nodes/* and server_routes
+# only), so shipping it would put files on thousands of machines that no
+# install ever reads.
+#
+# Everything here stays in the repository, where it is load-bearing: the tests
+# run in CI, and scripts/build_workflows.py generates workflows/ and is what a
+# contributor edits. Only the shipped package is smaller.
+#
+# workflows/ is deliberately NOT excluded — the templates are the product.
+
+tests/
+scripts/
+.github/
+
+# Tooling caches and editor droppings, in case a publish runs from a working
+# directory rather than a clean checkout.
+.pytest_cache/
+.ruff_cache/
+__pycache__/
+.DS_Store

styleref_nodes/apply.py

@@ -21,8 +21,7 @@ CATEGORY = "StyleRef"
 
 # Node-facing target → the server's `format` parameter.
 #
-# The list is EXACTLY the web app's copy-box format list (owner decision
-# 2026-07-22, superseding the earlier midjourney removal), in the same order:
+# The list is EXACTLY the web app's copy-box format list, in the same order:
 # AI Tools, STYLE.md, Flux, Midjourney, Diffusion, JSON. The style means the
 # same thing here as on styleref.io, format for format.
 #

styleref_nodes/facets.py

@@ -5,8 +5,8 @@ Apply is the paved road. This is the escape hatch: the style's sections come
 out as individual STRING outputs so any field can drive things a prompt string
 cannot — IPAdapter weights, conditioning regions, per-section text encoders.
 
-The output list is EXACTLY the style schema's section list (owner decision
-2026-07-22), in schema order, plus the six custom style items individually —
+The output list is EXACTLY the style schema's section list, in schema order,
+plus the six custom style items individually —
 nothing derived, nothing invented, so the node reads one-to-one against the
 style board on styleref.io. Output names are the schema's own section ids.
 The section list is data-driven from one constant: a future schema section is

tests/test_nodes.py

@@ -65,10 +65,10 @@ def test_pack_registers_exactly_the_five_nodes():
 
 def test_targets_are_exactly_the_web_apps_copy_list_in_order():
     """
-    Owner decision 2026-07-22: the target list mirrors the web
-    app's copy-box format list exactly — same entries, same order — so the
-    style means the same thing here as on styleref.io. Source of truth:
-    PROMPT_FORMAT_OPTIONS in src/components/ui/PromptFormatDropdown.tsx.
+    The target list mirrors the web app's copy-box format list exactly — same
+    entries, same order — so a style means the same thing here as it does on
+    styleref.io. The web app's list is the source of truth; this test is what
+    catches the two drifting apart.
     """
     assert list(apply_node.TARGETS) == [
         "ai_tools",
@@ -149,9 +149,8 @@ def test_composition_handles_an_empty_side():
 
 def test_negative_merges_guardrails_and_extra_only():
     """
-    No injected boilerplate (owner decision 2026-07-22): the style's guardrails
-    are the authority, and anything generic belongs in extra_negative where it
-    stays visible.
+    No injected boilerplate: the style's guardrails are the authority, and
+    anything generic belongs in extra_negative, where the user can see it.
     """
     negative = apply_node.compose_negative(STYLE, "extra thing")
     assert "lens flares" in negative
@@ -406,8 +405,8 @@ def test_load_returns_ui_text_and_result_tuple(monkeypatch):
 
 def test_load_never_fetches_style_text(monkeypatch):
     """
-    STYLE.md moved to Apply's target list (owner decision 2026-07-22): Load has
-    no style_doc output and must never spend the extra text-fetch call.
+    STYLE.md is a target on Apply, not an output on Load — so Load has no
+    style_doc output and must never spend the extra text-fetch call.
     """
     monkeypatch.setattr(
         api, "get_style_spec", lambda ref, etag=None: ({"name": "Warm", "sections": {}}, None, None)
@@ -532,9 +531,9 @@ def test_sections_filter_reaches_the_api_and_the_cache_key(monkeypatch):
 
 def test_facets_outputs_are_exactly_the_schema_sections_plus_custom_items():
     """
-    Owner decision 2026-07-22: Facets mirrors the style board one-to-one — the
-    schema's section list in schema order, plus the six custom style items,
-    and NOTHING derived (no palette/primary_color/width/height/negative).
+    Facets mirrors the style board one-to-one — the schema's section list in
+    schema order, plus the six custom style items, and NOTHING derived (no
+    palette/primary_color/width/height/negative).
     """
     from styleref_nodes import facets as facets_node