diff --git a/.nm-triage-nonprod b/.nm-triage-nonprod
new file mode 100644
index 0000000000..5409416478
--- /dev/null
+++ b/.nm-triage-nonprod
@@ -0,0 +1,11 @@
+# Paths that no-mistakes triage counts as non-production (one glob per line).
+# Conformance corpora and other test fixtures: data, gated by every port's runner.
+fixtures/**
+# The metamodel spec and the copies generated from it into each port.
+spec/metamodel/*.json
+server/csharp/MetaObjects/SpecMetamodel/*.json
+server/python/src/metaobjects/spec_metamodel/*.json
+server/typescript/packages/metadata/src/**/*.embedded.ts
+# The generated reference site and showcase payload.
+site-reference/**
+examples/showcase/site-payload.json
diff --git a/AGENTS.md b/AGENTS.md
index 7b73c724fe..3d772c9692 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -71,7 +71,7 @@ PyPI has had no product change since `0.25.0` — nothing is broken.
 - **Kotlin** — `codegen-kotlin` (KotlinPoet on JVM): entity + Exposed table + Spring controller + payload + relations + filter allowlist + validator + stored-proc + output-parser generators. `integration-tests-kotlin` runs the persistence-conformance corpus through Exposed against Testcontainers Postgres.
 
 **Cross-port conformance corpora** (every port runs the shared corpus):
-- Metamodel: `fixtures/conformance/` (458 fixtures; 38 shared corpora in total — per-corpus counts + the corpus x port matrix live in `docs/CONFORMANCE.md`). TS / C# / Java / Python all green.
+- Metamodel: `fixtures/conformance/` (459 fixtures; 38 shared corpora in total — per-corpus counts + the corpus x port matrix live in `docs/CONFORMANCE.md`). TS / C# / Java / Python all green.
 - Render: `fixtures/render-conformance/`. TS / C# / Java / Kotlin / Python byte-identical.
 - Persistence: `fixtures/persistence-conformance/`. **Query** scenarios run on every port (TS / C# / Java / Kotlin / Python), each provisioning its test DB by executing the committed, TS-produced `canonical/schema.postgres.sql` (Postgres only — Derby dropped for the cross-port query corpus, ADR-0015; the FR-044 `report-*` scenarios also run on SQLite, MySQL and D1 where a port has the engine, see [docs/CONFORMANCE.md](docs/CONFORMANCE.md)). The **migration** scenarios are exercised by **TS only** (TS owns schema migrations). **The corpus now gates WRITES, not just reads (SP-H):** an `op: roundtrip` scenario type INSERTs through each port's runtime/ORM write codec (NOT raw SQL), reads the row back, and asserts the wire-normalized value. The `AllTypes` entity (`roundtrip-all-types.yaml`) carries one field of **every** persistable `field.*` subtype — string/int/long/double/float/decimal/boolean/date/time/timestamp(+tz)/currency/enum/uuid/object — plus an **array-of-VO** `field.object @isArray @storage:jsonb` column (`labels`, written as 2-element / empty-`[]` / single-element arrays across the three rows) — so every subtype write+read (incl. the array-of-value-object jsonb codec) round-trips through every port against Testcontainers PG. (`field.byte`/`field.short`/`field.class` were cut as non-functional registration-only stubs — the matrix tracks only genuinely-supported subtypes; see `fixtures/registry-conformance/README.md` → "Per-subtype write-round-trip matrix".)
 - API-contract: `fixtures/api-contract-conformance/`. TS / C# / Java / Kotlin / Python all green — each port runs **two lanes**: a hand-rolled reference server AND its **generated** API artifact booted over HTTP (the deployed controller/routes; TS+C# full-stack vs Testcontainers PG, Java/Kotlin/Python generated controller + in-memory repo behind the consumer seam). The generated fan-out found 10 real deployment bugs golden snapshots missed. Three sub-corpora run the **generated lane only, on all five ports** — `write-through/`, `projection/` (F22: a view-only `object.projection` serves GET list + GET by id and answers every write verb with `405 {"error": "method_not_allowed"}`) and `report/` (FR-044: a view-backed `object.report` serves GET list, answers `POST` with that same `405`, and mounts no `/{id}` route). That is deliberate, not a gap: what is under test is whether a port's GENERATOR emits those routes, and a hand-rolled reference server would answer every scenario by construction. The `m2m/` sub-corpus also gates **TPH x M:N together** (base-declared, subtype-declared, abstract-mid-declared, a non-subtype source onto a subtype TARGET, and the cross-subtype source id answering `200 []`) — the two corpora were originally built disjoint (`tph/` had no relationships, `m2m/` no discriminators), which is precisely how that defect class survived.
diff --git a/CHANGELOG.md b/CHANGELOG.md
index dc0a6175ac..5e2d310b29 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -10,6 +10,27 @@ here.**
 
 ## [Unreleased]
 
+### Fixed
+
+- **A projection view keeps the columns a reverse `@via` hop feeds.** An origin `@via` that walks
+  an `identity.reference` backwards (`Project.timeEntries`, #560) loaded, but `meta migrate` left
+  its `origin.aggregate` / `origin.first` columns out of `CREATE VIEW` without a word. The view
+  emitter now resolves the hop the way the loader does and joins from the holder of the
+  reference. Gated by a Postgres test that creates the view and reads it.
+- **`origin.aggregate` `sum`/`avg`/`min`/`max` over a value-object field is refused at load.**
+  `@of: Invoice.total` on a flattened Money field loaded and emitted `SUM(i.total)`, a column that
+  does not exist. Every port now refuses it with `ERR_INVALID_ORIGIN` and says to aggregate a
+  scalar field; gated by the `error-origin-aggregate-value-object` conformance fixture. Refusing,
+  rather than summing the flattened amount column, keeps a sum from adding amounts in different
+  currencies. `count` is unaffected.
+
+### Changed
+
+- **Agent context routes report work to the reporting vocabulary.** The `meta init` agent guide
+  and the `metaobjects-authoring` skill now send grouped totals, dashboard numbers and read-only
+  report views to `object.report` (`references/declaring-reports.md`) before the `@sql` escape,
+  which they describe as the last resort for bodies the vocabulary cannot express.
+
 ## [1.3.0] — 2026-11-11
 
 _npm `1.3.0` (full lockstep across all 14 `@metaobjectsdev/*` publish candidates)._
diff --git a/agent-context/skills/metaobjects-authoring/SKILL.md b/agent-context/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/agent-context/skills/metaobjects-authoring/SKILL.md
+++ b/agent-context/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/agent-context/skills/metaobjects-authoring/references/read-views-and-projections.md b/agent-context/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/agent-context/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/agent-context/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/agent-context/skills/metaobjects-authoring/references/sources.md b/agent-context/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/agent-context/skills/metaobjects-authoring/references/sources.md
+++ b/agent-context/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/agent-context/templates/always-on.md.mustache b/agent-context/templates/always-on.md.mustache
index cc0003da65..cc1aaaf1bf 100644
--- a/agent-context/templates/always-on.md.mustache
+++ b/agent-context/templates/always-on.md.mustache
@@ -50,6 +50,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/docs/CONFORMANCE.md b/docs/CONFORMANCE.md
index 4fac400837..24b413c16c 100644
--- a/docs/CONFORMANCE.md
+++ b/docs/CONFORMANCE.md
@@ -26,7 +26,7 @@ regenerate with `ls -d fixtures/<corpus>/*/ | wc -l` for directory-shaped corpor
 
 | Corpus | Fixtures | TS | Java | Kotlin | C# | Python |
 |---|---|---|---|---|---|---|
-| [`fixtures/conformance/`](../fixtures/conformance/) (metamodel) | 458 | ✓ | ✓ | inherits via `metadata-ktx` | ✓ | ✓ |
+| [`fixtures/conformance/`](../fixtures/conformance/) (metamodel) | 459 | ✓ | ✓ | inherits via `metadata-ktx` | ✓ | ✓ |
 | [`fixtures/yaml-conformance/`](../fixtures/yaml-conformance/) | 16 | 16 / 16 | 15 / 16 (1 ledgered: `yaml-quoted-leading-zero` — Java pipeline strips quotes off `"007"`) | inherits via Java | 15 / 16 (1 ledgered: `error-yaml-coerced-hex-in-string` — YamlDotNet doesn't coerce `0xFF`) | 16 / 16 |
 | [`fixtures/verify-conformance/`](../fixtures/verify-conformance/) | 45 | ✓ | ✓ | inherits via Java | ✓ | ✓ |
 | [`fixtures/verify-strict-conformance/`](../fixtures/verify-strict-conformance/) | 1 | ✓ | — | — | — | ✓ |
@@ -354,7 +354,7 @@ unit-test runners (`bun test`, `dotnet test`, `pytest`, `mvn test`) pull Docker.
 
 ## Fixture-to-doc mapping
 
-### `fixtures/conformance/` — metamodel loader + canonical serializer (458)
+### `fixtures/conformance/` — metamodel loader + canonical serializer (459)
 
 | Fixture prefix | Feature doc |
 |---|---|
@@ -716,7 +716,7 @@ JSON and verify payload feeds it unchanged.
 
 ## Orphaned fixtures (tested but not yet documented)
 
-The fixtures in the nine corpora mapped above (metamodel 458 + yaml 16 + verify 45
+The fixtures in the nine corpora mapped above (metamodel 459 + yaml 16 + verify 45
 + render 18 + persistence 62 + api-contract 123 + source-resolution 25 + scope 10 +
 dependency 47) each map to a feature doc, and so do the two requirement corpora, `shared-model-conformance`, the
 `cube-model` corpus and `explorer-conformance`, whose case lists live in their own READMEs. None are orphaned today. The remaining
diff --git a/docs/features/source-kinds.md b/docs/features/source-kinds.md
index 6840c71a10..3fe738009d 100644
--- a/docs/features/source-kinds.md
+++ b/docs/features/source-kinds.md
@@ -182,6 +182,12 @@ max over only the active rows — the filtered-scalar shape a plain aggregate ca
 ]}}
 ```
 
+**`sum` / `avg` / `min` / `max` need a scalar `@of`.** A value-object field (`field.object`,
+flattened or jsonb) is several columns or a document, so there is no one column to reduce, and a
+Money-shaped sum would add amounts in different currencies. The loader refuses it in every port
+(`ERR_INVALID_ORIGIN`); model the amount as its own scalar field and group by its currency.
+`count` still counts rows.
+
 **More `origin.aggregate` reductions (#195).** Beyond the scalar reduces, `@agg` also takes
 `any` / `all` (predicate quantifiers over a `@filter` — `@of` forbidden; empty set →
 `any=false`, `all=true`) and `collect` (an array rollup of `@of` into an `isArray` field,
diff --git a/examples/showcase/site-payload.json b/examples/showcase/site-payload.json
index 78478a6803..c12759239a 100644
--- a/examples/showcase/site-payload.json
+++ b/examples/showcase/site-payload.json
@@ -7,7 +7,7 @@
     "metamodel": "1.3"
   },
   "counts": {
-    "fixtures": 458,
+    "fixtures": 459,
     "corpora": 38,
     "baseTypes": 17
   },
diff --git a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/SKILL.md b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/SKILL.md
+++ b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/sources.md b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/sources.md
+++ b/fixtures/agent-context-conformance/csharp/expected/.claude/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/fixtures/agent-context-conformance/csharp/expected/.metaobjects/AGENTS.md b/fixtures/agent-context-conformance/csharp/expected/.metaobjects/AGENTS.md
index cc76b4b873..a3f4014b0b 100644
--- a/fixtures/agent-context-conformance/csharp/expected/.metaobjects/AGENTS.md
+++ b/fixtures/agent-context-conformance/csharp/expected/.metaobjects/AGENTS.md
@@ -68,6 +68,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/csharp/expected/.metaobjects/CLAUDE.md b/fixtures/agent-context-conformance/csharp/expected/.metaobjects/CLAUDE.md
index cc76b4b873..a3f4014b0b 100644
--- a/fixtures/agent-context-conformance/csharp/expected/.metaobjects/CLAUDE.md
+++ b/fixtures/agent-context-conformance/csharp/expected/.metaobjects/CLAUDE.md
@@ -68,6 +68,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
+++ b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
+++ b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/AGENTS.md b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/AGENTS.md
index b700a9d00e..6c4200f7c4 100644
--- a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/AGENTS.md
+++ b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/AGENTS.md
@@ -87,6 +87,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/CLAUDE.md b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/CLAUDE.md
index b700a9d00e..6c4200f7c4 100644
--- a/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/CLAUDE.md
+++ b/fixtures/agent-context-conformance/java-kotlin-react-tanstack/expected/.metaobjects/CLAUDE.md
@@ -87,6 +87,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/SKILL.md b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/SKILL.md
+++ b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/sources.md b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/sources.md
+++ b/fixtures/agent-context-conformance/java-react/expected/.claude/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/fixtures/agent-context-conformance/java-react/expected/.metaobjects/AGENTS.md b/fixtures/agent-context-conformance/java-react/expected/.metaobjects/AGENTS.md
index 26abb81c10..074f87fdd1 100644
--- a/fixtures/agent-context-conformance/java-react/expected/.metaobjects/AGENTS.md
+++ b/fixtures/agent-context-conformance/java-react/expected/.metaobjects/AGENTS.md
@@ -73,6 +73,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/java-react/expected/.metaobjects/CLAUDE.md b/fixtures/agent-context-conformance/java-react/expected/.metaobjects/CLAUDE.md
index 26abb81c10..074f87fdd1 100644
--- a/fixtures/agent-context-conformance/java-react/expected/.metaobjects/CLAUDE.md
+++ b/fixtures/agent-context-conformance/java-react/expected/.metaobjects/CLAUDE.md
@@ -73,6 +73,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/SKILL.md b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/SKILL.md
+++ b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/sources.md b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/sources.md
+++ b/fixtures/agent-context-conformance/python/expected/.claude/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/fixtures/agent-context-conformance/python/expected/.metaobjects/AGENTS.md b/fixtures/agent-context-conformance/python/expected/.metaobjects/AGENTS.md
index 8f925a267a..4ac33cdba1 100644
--- a/fixtures/agent-context-conformance/python/expected/.metaobjects/AGENTS.md
+++ b/fixtures/agent-context-conformance/python/expected/.metaobjects/AGENTS.md
@@ -69,6 +69,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/python/expected/.metaobjects/CLAUDE.md b/fixtures/agent-context-conformance/python/expected/.metaobjects/CLAUDE.md
index 8f925a267a..4ac33cdba1 100644
--- a/fixtures/agent-context-conformance/python/expected/.metaobjects/CLAUDE.md
+++ b/fixtures/agent-context-conformance/python/expected/.metaobjects/CLAUDE.md
@@ -69,6 +69,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
index de11bce2cc..459023b776 100644
--- a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
+++ b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/SKILL.md
@@ -31,10 +31,10 @@ This file covers what almost every model needs. The topics below live in
 | `references/identities-and-indexes.md` | declaring a primary, secondary or reference identity, an FK, or a non-unique `index.lookup` |
 | `references/sources.md` | the `source.rdb` shape and `@kind`; a store MetaObjects does not manage; a derived or computed column (reach for an entity read-view first) |
 | `references/abstracts-and-overlay.md` | a shared `abstract` base, `extends`, or a layered `overlay` file |
-| `references/declaring-reports.md` | declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
+| `references/declaring-reports.md` | **open first** for a report, dashboard, totals, `GROUP BY`, "per client / per month" or a read-only summary view: declaring `dimension.*`, `measure.*`, `segment.filter` or an `object.report` |
 | `references/model-first-rationale.md` | someone argues that hand-writing a layer is cheaper this once |
 | `references/relationships.md` | a many-to-many link or self-join, or a "rows that reference this one" query |
-| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view |
+| `references/read-views-and-projections.md` | an `object.projection`, `origin.*` vocabulary, `@filter` / `@expr`, or an `@sql` / `@unmanaged` view (a grouped or aggregate view is `declaring-reports.md` first) |
 | `references/inheritance-tph.md` | several entities are variants of one thing sharing a single table (`@discriminator`) |
 | `references/metadata-dependencies.md` | the project builds on another package's metadata (`dependencies`, cross-package `overlay`) |
 | `references/reporting.md` | what a declared report gives you: what columns a report gets, time grains, null rules, rows with no facts (`@spine`), defaults (`@default`), engine differences |
diff --git a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
index 06d2ccf2e8..99422ce11f 100644
--- a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
+++ b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/read-views-and-projections.md
@@ -1,6 +1,6 @@
 # Projections, `origin.*`, and managed views
 
-> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view.
+> Part of the `metaobjects-authoring` skill. The skill covers the entity read-view, which is the right answer for most grids. Read this when you need an independent exposure (`object.projection`), the `origin.*` vocabulary, `@filter` / `@expr` shapes, or an `@sql` / `@unmanaged` view. A grouped total or report (per client, per month) is an `object.report`: `declaring-reports.md`.
 
 ## Contents
 - Read-view or projection?
@@ -125,8 +125,9 @@ non-aggregate-derived** projection fields: naming an undeclared field fails the
 
 **The `CREATE VIEW` body is generated from those `origin.*` children by the Node
 `meta migrate` — never hand-author view SQL for a shape origins can express** (an unmodeled
-view is *unmanaged*, so `meta verify --db` can't even catch the drift). For a genuinely
-irreducible body (recursive CTE, window function, set operation) that origins can't express,
+view is *unmanaged*, so `meta verify --db` can't even catch the drift). **An `@sql` body that
+`GROUP BY`s: check `declaring-reports.md` first** (`object.report`, `@spine`, `@default`).
+For a genuinely irreducible body (recursive CTE, window function, set operation) that origins can't express,
 carry it in the `source.rdb` **`@sql`** escape (#208, ADR-0043) — a hand-written body the
 tool registers, fingerprints, and drift-checks (adopt a pre-existing view with
 `meta migrate --allow adopt-view`) — rather than a hand-edited migration file where it goes
diff --git a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
index 5f7f94c031..bff437f410 100644
--- a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
+++ b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.claude/skills/metaobjects-authoring/references/sources.md
@@ -94,4 +94,6 @@ column because the filter tier runs against the view.
 **When you need an independent exposure** — a subset, renamed columns, a row-filtered
 view, a versioned or external shape — that is an `object.projection`. It, the `origin.*`
 vocabulary, the structured `@filter` / `@expr` shapes and the `@sql` / `@unmanaged` escapes
-are in `references/read-views-and-projections.md`.
+are in `references/read-views-and-projections.md`. A grouped or aggregate view (totals per client
+or per month) is an `object.report` (`references/declaring-reports.md`); `@sql` is for what neither
+can express (a recursive CTE, a window function, a set operation).
diff --git a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/AGENTS.md b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/AGENTS.md
index f71ff6288e..32ec038942 100644
--- a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/AGENTS.md
+++ b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/AGENTS.md
@@ -86,6 +86,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/CLAUDE.md b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/CLAUDE.md
index f71ff6288e..32ec038942 100644
--- a/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/CLAUDE.md
+++ b/fixtures/agent-context-conformance/ts-react-tanstack/expected/.metaobjects/CLAUDE.md
@@ -86,6 +86,9 @@ Commit the metadata, the generated code, `.metaobjects/migrations/` and
 - Once the project declares a `template.prompt`, every LLM prompt is one: the text in a
   template, the data in a payload, a structured reply read through its generated extractor
   (`metaobjects-prompts`).
+- A grouped total, dashboard number or read-only report view (per client, per month) is an
+  `object.report` over `dimension.*` / `measure.*` (`metaobjects-authoring`,
+  `references/declaring-reports.md`). A hand-written `@sql` view body is the last resort.
 - Never invent metadata vocabulary. Search it: `npx meta types <term>`, and
   `npx meta types <type>.<subType> --detail` for one construct's valid attributes.
 - Metadata nodes are fused-key maps, `{"field.string": {"name": "email"}}`. Attribute names are
diff --git a/fixtures/conformance/error-origin-aggregate-value-object/expected-errors.json b/fixtures/conformance/error-origin-aggregate-value-object/expected-errors.json
new file mode 100644
index 0000000000..3290a6cb63
--- /dev/null
+++ b/fixtures/conformance/error-origin-aggregate-value-object/expected-errors.json
@@ -0,0 +1,17 @@
+{
+  "errors": [
+    {
+      "code": "ERR_INVALID_ORIGIN",
+      "source": {
+        "format": "resolved",
+        "files": [
+          "meta.billing.json"
+        ],
+        "jsonPath": "$['metadata.root'].children[3]['object.projection'].children[2]['field.decimal'].children[0]['origin.aggregate']",
+        "referrer": "ProjectInvoicing::invoicedTotal",
+        "target": "Invoice.total"
+      }
+    }
+  ],
+  "warnings": []
+}
diff --git a/fixtures/conformance/error-origin-aggregate-value-object/input/meta.billing.json b/fixtures/conformance/error-origin-aggregate-value-object/input/meta.billing.json
new file mode 100644
index 0000000000..4f13778018
--- /dev/null
+++ b/fixtures/conformance/error-origin-aggregate-value-object/input/meta.billing.json
@@ -0,0 +1,58 @@
+{
+  "metadata.root": {
+    "package": "acme::billing",
+    "children": [
+      {
+        "object.value": {
+          "name": "Money",
+          "children": [
+            { "field.decimal": { "name": "amount", "@precision": 12, "@scale": 2 } },
+            { "field.string": { "name": "currency", "@maxLength": 3 } }
+          ]
+        }
+      },
+      {
+        "object.entity": {
+          "name": "Project",
+          "children": [
+            { "source.rdb": { "@table": "projects" } },
+            { "field.long": { "name": "id" } },
+            { "identity.primary": { "name": "id", "@fields": "id" } },
+            { "relationship.association": { "name": "invoices", "@objectRef": "Invoice", "@cardinality": "many" } }
+          ]
+        }
+      },
+      {
+        "object.entity": {
+          "name": "Invoice",
+          "children": [
+            { "source.rdb": { "@table": "invoices" } },
+            { "field.long": { "name": "id" } },
+            { "field.long": { "name": "projectId" } },
+            { "field.object": { "name": "total", "@objectRef": "Money", "@storage": "flattened" } },
+            { "identity.primary": { "name": "id", "@fields": "id" } },
+            { "identity.reference": { "name": "project", "@fields": "projectId", "@references": "Project" } }
+          ]
+        }
+      },
+      {
+        "object.projection": {
+          "name": "ProjectInvoicing",
+          "children": [
+            { "source.rdb": { "@kind": "view", "@view": "v_project_invoicing" } },
+            { "field.long": { "name": "id", "extends": "Project.id" } },
+            {
+              "field.decimal": {
+                "name": "invoicedTotal",
+                "children": [
+                  { "origin.aggregate": { "@agg": "sum", "@of": "Invoice.total", "@via": "Project.invoices" } }
+                ]
+              }
+            },
+            { "identity.primary": { "name": "id", "extends": "Project.id" } }
+          ]
+        }
+      }
+    ]
+  }
+}
diff --git a/fixtures/conformance/error-origin-aggregate-value-object/providers.json b/fixtures/conformance/error-origin-aggregate-value-object/providers.json
new file mode 100644
index 0000000000..f9bb435a37
--- /dev/null
+++ b/fixtures/conformance/error-origin-aggregate-value-object/providers.json
@@ -0,0 +1 @@
+["metaobjects-core-types","metaobjects-db"]
diff --git a/server/csharp/MetaObjects/Loader/ValidationPasses.cs b/server/csharp/MetaObjects/Loader/ValidationPasses.cs
index 04612bb0c5..c9fe25c3db 100644
--- a/server/csharp/MetaObjects/Loader/ValidationPasses.cs
+++ b/server/csharp/MetaObjects/Loader/ValidationPasses.cs
@@ -703,6 +703,20 @@ public static IReadOnlyList<MetaError> ValidateOriginPaths(MetaData root)
                         // NOTE (FR-024 B6): NO extends/origin agreement on aggregates.
                         var ofTarget = ValidateFromPath(of, root, obj, field.Name, errors,
                             "origin.aggregate.@of", src);
+                        // A sum/avg/min/max reduces ONE column. A value-object field is several columns
+                        // (flattened) or a document (jsonb), so there is no column to reduce, and summing a
+                        // Money-shaped value would add amounts in different currencies.
+                        // ADR-0039: resolving — @objectRef may be inherited via extends.
+                        if (ofTarget is ResolvedFromTarget valueTarget
+                            && agg is not null && Array.IndexOf(COLUMN_VALUE_AGGREGATES, agg) >= 0
+                            && (valueTarget.Field.SubType == FIELD_SUBTYPE_OBJECT
+                                || valueTarget.Field.Attr(FIELD_ATTR_OBJECT_REF) is not null))
+                            errors.Add(new MetaError(
+                                $"origin.aggregate @agg:{agg} on {obj.Name}.{field.Name}: @of \"{of}\" is a value-object " +
+                                $"field (field.{valueTarget.Field.SubType}), which has no single column to aggregate. " +
+                                "Aggregate a scalar field instead: model the amount as its own field (and group by its currency).",
+                                ErrorCode.ERR_INVALID_ORIGIN,
+                                Envelope: ResolvedSource.From(src, $"{obj.Fqn()}::{field.Name}", of)));
                         // #195 — collect preserves the element type: the array field's own
                         // subType must equal the @of column's subType (the #185 doctrine).
                         if (isCollect && ofTarget is ResolvedFromTarget collectTarget
diff --git a/server/csharp/MetaObjects/Persistence/Origin/OriginConstants.cs b/server/csharp/MetaObjects/Persistence/Origin/OriginConstants.cs
index 9d57eb3d71..42bb13f82a 100644
--- a/server/csharp/MetaObjects/Persistence/Origin/OriginConstants.cs
+++ b/server/csharp/MetaObjects/Persistence/Origin/OriginConstants.cs
@@ -105,6 +105,11 @@ public static class OriginConstants
     public const string AGG_ALL     = "all";
     public const string AGG_COLLECT = "collect";
 
+    /// <summary>The @agg values that reduce the VALUES of the @of column (count only counts
+    /// non-null rows). Mirrors the TS COLUMN_VALUE_AGGREGATES.</summary>
+    public static readonly string[] COLUMN_VALUE_AGGREGATES =
+        [AGGREGATE_FN_SUM, AGGREGATE_FN_AVG, AGGREGATE_FN_MIN, AGGREGATE_FN_MAX];
+
     /// <summary>The full @agg attribute vocabulary (matches origin.json allowedValues).</summary>
     public static readonly string[] ORIGIN_AGG_VALUES =
         [AGGREGATE_FN_COUNT, AGGREGATE_FN_SUM, AGGREGATE_FN_AVG, AGGREGATE_FN_MIN, AGGREGATE_FN_MAX,
diff --git a/server/java/metadata/src/main/java/com/metaobjects/loader/ValidationPhase.java b/server/java/metadata/src/main/java/com/metaobjects/loader/ValidationPhase.java
index b26e8eeed1..cd3f2fd32c 100644
--- a/server/java/metadata/src/main/java/com/metaobjects/loader/ValidationPhase.java
+++ b/server/java/metadata/src/main/java/com/metaobjects/loader/ValidationPhase.java
@@ -3114,6 +3114,23 @@ private static void validateOriginNode(MetaRoot root, MetaObject obj,
             }
             OriginTarget ofTarget = validateFromOrOfPath(of, root, obj, field.getName(),
                 "origin.aggregate.@of", src);
+            // A sum/avg/min/max reduces ONE column. A value-object field is several columns
+            // (flattened) or a document (jsonb), so there is no column to reduce, and summing a
+            // Money-shaped value would add amounts in different currencies.
+            // ADR-0039: resolving (hasMetaAttr defaults to includeParentData) — @objectRef may be
+            // inherited via extends.
+            if (agg != null && MetaOrigin.COLUMN_VALUE_AGGREGATES.contains(agg)
+                    && (ObjectField.SUBTYPE_OBJECT.equals(ofTarget.field.getSubType())
+                        || ofTarget.field.hasMetaAttr(ObjectField.ATTR_OBJECTREF))) {
+                throw new MetaDataException(
+                    ErrorMessageConstants.ERR_INVALID_ORIGIN
+                        + ": origin.aggregate @agg:" + agg + " on " + obj.getName() + "." + field.getName()
+                        + ": @of \"" + of + "\" is a value-object field (field." + ofTarget.field.getSubType()
+                        + "), which has no single column to aggregate. Aggregate a scalar field instead:"
+                        + " model the amount as its own field (and group by its currency).",
+                    ErrorCode.ERR_INVALID_ORIGIN,
+                    ResolvedSource.from(src, obj.getShortName() + "::" + field.getName(), of));
+            }
             // #195 — collect preserves the element type: the array field's own subType
             // must equal the @of column's subType (the #185 doctrine on the element).
             if (isCollect && !field.getSubType().equals(ofTarget.field.getSubType())) {
diff --git a/server/java/metadata/src/main/java/com/metaobjects/origin/MetaOrigin.java b/server/java/metadata/src/main/java/com/metaobjects/origin/MetaOrigin.java
index c0d9a5771f..ab238a77e5 100644
--- a/server/java/metadata/src/main/java/com/metaobjects/origin/MetaOrigin.java
+++ b/server/java/metadata/src/main/java/com/metaobjects/origin/MetaOrigin.java
@@ -124,6 +124,11 @@ public abstract class MetaOrigin extends MetaData {
     public static final String AGG_ALL     = "all";
     public static final String AGG_COLLECT = "collect";
 
+    /** The {@code @agg} values that reduce the VALUES of the {@code @of} column (count only counts
+     *  non-null rows). Mirrors the TS COLUMN_VALUE_AGGREGATES. */
+    public static final java.util.List<String> COLUMN_VALUE_AGGREGATES =
+        java.util.List.of(AGG_SUM, AGG_AVG, AGG_MIN, AGG_MAX);
+
     /**
      * Closed set of valid {@code @agg} values. Used by
      * {@link com.metaobjects.loader.ValidationPhase} for enum-membership checks.
diff --git a/server/python/src/metaobjects/loader/validation_passes.py b/server/python/src/metaobjects/loader/validation_passes.py
index d3c566cb96..54280e7b85 100644
--- a/server/python/src/metaobjects/loader/validation_passes.py
+++ b/server/python/src/metaobjects/loader/validation_passes.py
@@ -105,6 +105,7 @@
     AGG_ANY,
     AGG_COLLECT,
     ASSEMBLY_ORIGIN_SUBTYPES,
+    COLUMN_VALUE_AGGREGATES,
     ORIGIN_ATTR_AGG,
     ORIGIN_ATTR_CONVERT,
     ORIGIN_ATTR_DISTINCT,
@@ -2500,6 +2501,23 @@ def _validate_origin_paths(
                 of_target = _validate_entity_field_ref(
                     of_ref, ORIGIN_ATTR_OF, ctx, root, host_pkg, errors, origin, referrer
                 )
+                # A sum/avg/min/max reduces ONE column. A value-object field is several
+                # columns (flattened) or a document (jsonb), so there is no column to reduce,
+                # and summing a Money-shaped value would add amounts in different currencies.
+                # Resolving read: @objectRef may be inherited via extends (ADR-0039).
+                if (
+                    of_target is not None
+                    and agg in COLUMN_VALUE_AGGREGATES
+                    and (of_target[1].sub_type == FIELD_SUBTYPE_OBJECT
+                         or of_target[1].get_meta_attr(FIELD_ATTR_OBJECT_REF) is not None)
+                ):
+                    errors.append(MetaError(
+                        f"origin.aggregate @agg:{agg} on {obj.name}.{node.name}: @of \"{of_ref}\" is a "
+                        f"value-object field (field.{of_target[1].sub_type}), which has no single column "
+                        f"to aggregate. Aggregate a scalar field instead: model the amount as its own "
+                        f"field (and group by its currency).",
+                        ErrorCode.ERR_INVALID_ORIGIN,
+                        envelope=resolved_source(origin.source, referrer, of_ref)))
                 # #195 — collect preserves the element type: the array field's own subType
                 # must equal the @of column's subType (the #185 doctrine on the element).
                 if is_collect and of_target is not None and node.sub_type != of_target[1].sub_type:
diff --git a/server/python/src/metaobjects/meta/persistence/origin/origin_constants.py b/server/python/src/metaobjects/meta/persistence/origin/origin_constants.py
index 6097df59c5..04d95b23f7 100644
--- a/server/python/src/metaobjects/meta/persistence/origin/origin_constants.py
+++ b/server/python/src/metaobjects/meta/persistence/origin/origin_constants.py
@@ -51,6 +51,9 @@
 AGG_ANY = "any"
 AGG_ALL = "all"
 AGG_COLLECT = "collect"
+# The @agg values that reduce the VALUES of the @of column (count only counts
+# non-null rows). Mirrors the TS COLUMN_VALUE_AGGREGATES.
+COLUMN_VALUE_AGGREGATES = ("sum", "avg", "min", "max")
 ORIGIN_ATTR_OF = "of"
 # @filter (attr.filter object): an optional PORTABLE structured predicate scoping the
 # aggregated rows (same shape as a preset filter). On @agg:any|all it is the QUANTIFIED
diff --git a/server/typescript/packages/codegen-ts/src/projection/extract-view-spec.ts b/server/typescript/packages/codegen-ts/src/projection/extract-view-spec.ts
index 8fd32e963d..639c83b7ee 100644
--- a/server/typescript/packages/codegen-ts/src/projection/extract-view-spec.ts
+++ b/server/typescript/packages/codegen-ts/src/projection/extract-view-spec.ts
@@ -54,6 +54,7 @@ import {
   FIELD_ATTR_COLUMN,
   OBJECT_PROJECTION_ATTR_FILTER,
   findReferencesBetween,
+  findReverseReferenceHops,
   resolveRelationshipReference,
   resolveObjectRef,
   refTarget,
@@ -353,11 +354,16 @@ function findReferenceHop(obj: MetaData, name: string): MetaData | undefined {
 }
 
 /** Resolve a `@via` hop (relationship OR reference) to its target + cardinality.
- *  A reference hop is a to-one forward FK; its target is `@references`. */
+ *  A reference hop is a to-one forward FK; its target is `@references`. A hop naming
+ *  neither may walk an identity.reference BACKWARDS (#560): `Project.timeEntries` is the
+ *  reference TimeEntry holds onto Project, named by TimeEntry's collection name — a to-many
+ *  hop whose target is the holder (`reverse`). It resolves only when exactly one reference
+ *  answers the name; the loader refuses an ambiguous one (ERR_AMBIGUOUS_PATH). */
 function resolveHop(
   obj: MetaData,
   name: string,
-): { hop: MetaData; target: MetaObject | undefined; cardinality: "one" | "many" } | undefined {
+  root: MetaRoot,
+): { hop: MetaData; target: MetaObject | undefined; cardinality: "one" | "many"; reverse?: true } | undefined {
   // ADR-0059 R3/R7 — a hop's target is the object its reference attribute binds to, resolved
   // in the package of the node that declares that attribute, never the walked entity's.
   const rel = findRelationship(obj, name);
@@ -377,7 +383,11 @@ function resolveHop(
     const target = refTarget(ref, IDENTITY_REFERENCE_ATTR_REFERENCES);
     return { hop: ref, target: isMetaObject(target) ? target : undefined, cardinality: "one" };
   }
-  return undefined;
+  if (!isMetaObject(obj)) return undefined;
+  const reverse = findReverseReferenceHops(obj, name, root.objects());
+  if (reverse.length !== 1) return undefined;
+  const { holder, referenceIdentity } = reverse[0]!;
+  return { hop: referenceIdentity, target: holder, cardinality: "many", reverse: true };
 }
 
 /**
@@ -404,7 +414,10 @@ function resolveHopReference(
   hop: MetaData,
   hopName: string,
   target: MetaObject,
+  reverse = false,
 ): ReferenceLookup | ReferenceLookup[] {
+  // #560 — a reverse hop walks the reference `target` holds onto `holder`.
+  if (reverse) return { holder: target, other: holder, referenceIdentity: hop as unknown as MetaReferenceIdentity };
   if (hop.type === TYPE_IDENTITY && hop.subType === IDENTITY_SUBTYPE_REFERENCE) {
     return { holder, other: target, referenceIdentity: hop as unknown as MetaReferenceIdentity };
   }
@@ -424,11 +437,11 @@ function resolveHopReference(
  * reads it for the reference's full `@fields` list, where a view join reads the first.
  */
 export function hopReferenceIdentity(holder: MetaObject, hopName: string, root: MetaRoot): MetaReferenceIdentity | undefined {
-  const resolved = resolveHop(holder, hopName);
+  const resolved = resolveHop(holder, hopName, root);
   if (!resolved) return undefined;
   const target = resolved.target;
   if (!target) return undefined;
-  const ref = resolveHopReference(holder, resolved.hop, hopName, target);
+  const ref = resolveHopReference(holder, resolved.hop, hopName, target, resolved.reverse);
   if (Array.isArray(ref)) return ref.length === 1 ? ref[0]!.referenceIdentity : undefined;
   return ref.referenceIdentity;
 }
@@ -705,7 +718,7 @@ function viaTerminalEntity(via: string, root: MetaRoot, referrerPkg: string): Me
   let currentObj = resolveEntityRef(root, rawEntity, referrerPkg);
   if (!currentObj) return undefined;
   for (const relName of segments.slice(1)) {
-    const resolved = resolveHop(currentObj, relName);
+    const resolved = resolveHop(currentObj, relName, root);
     if (!resolved) return undefined;
     const target = resolved.target;
     if (!target) return undefined;
@@ -858,9 +871,9 @@ export function walkViaPath(via: string, root: MetaRoot, referrerPkg: string, ct
     // FR-024: a hop may name a relationship OR a reference-only FK
     // (identity.reference — a to-one forward-FK edge). ADR-0039: resolving —
     // a traversed relationship/reference may inherit its target via extends.
-    const resolved = resolveHop(currentObj, relName);
+    const resolved = resolveHop(currentObj, relName, root);
     if (!resolved) break;
-    const { hop, target, cardinality } = resolved;
+    const { hop, target, cardinality, reverse } = resolved;
     if (!target) break;
 
     // #368: two identity.reference declarations onto the same target are legal
@@ -869,7 +882,7 @@ export function walkViaPath(via: string, root: MetaRoot, referrerPkg: string, ct
     // one from the target alone, so an explicit `@via: "Match.homeTeamRef"` (or a
     // relationship disambiguated by @sourceRefField/name-pairing) resolves cleanly.
     // Only a relationship hop that even the ladder cannot choose reaches the throw.
-    const resolvedRef = resolveHopReference(currentObj as MetaObject, hop, relName, target);
+    const resolvedRef = resolveHopReference(currentObj as MetaObject, hop, relName, target, reverse);
     let ref: ReferenceLookup | undefined;
     if (Array.isArray(resolvedRef)) {
       if (resolvedRef.length > 1) {
@@ -919,8 +932,9 @@ export function walkViaPath(via: string, root: MetaRoot, referrerPkg: string, ct
 
     const resolvedPkField = ref.referenceIdentity.resolvedTargetPkField(root) ?? "id";
 
+    // A reverse hop's reference is always on the far side, even on a self-reference.
     const referenceHolder: "source" | "target" =
-      ref.holder.name === currentObj.name ? "source" : "target";
+      !reverse && ref.holder.name === currentObj.name ? "source" : "target";
 
     // FK lives on the holder; PK on the entity it references. Resolve both to
     // physical columns now so the ON clause is naming-strategy correct.
diff --git a/server/typescript/packages/integration-tests/test/origin-reverse-hop-view-pg.test.ts b/server/typescript/packages/integration-tests/test/origin-reverse-hop-view-pg.test.ts
new file mode 100644
index 0000000000..ddc738b7f7
--- /dev/null
+++ b/server/typescript/packages/integration-tests/test/origin-reverse-hop-view-pg.test.ts
@@ -0,0 +1,99 @@
+/**
+ * Postgres: a projection view whose origin `@via` walks an identity.reference BACKWARDS (#560).
+ *
+ * `Project.timeEntries` names no relationship: it walks TimeEntry's `project` reference from the
+ * entity it targets back to the entity holding it. The loader accepted that hop, but the view
+ * emitter resolved hops only through relationships and forward references, so it silently left
+ * the aggregate and first columns out of CREATE VIEW. This test migrates an empty database, puts
+ * rows in, and reads the view: a missing column, or a join pointed the wrong way, fails it.
+ */
+import { describe, test, expect, beforeAll, afterAll } from "bun:test";
+import { buildExpectedSchema, collectUnmanagedNames, diff, emit, introspectPostgres, splitSqlStatements } from "@metaobjectsdev/migrate-ts";
+import { buildProjectionViews } from "@metaobjectsdev/codegen-ts";
+import { MetaDataLoader, InMemoryStringSource, type MetaRoot } from "@metaobjectsdev/metadata";
+import { Kysely, PostgresDialect, sql } from "kysely";
+import { Pool } from "pg";
+import { startPostgres, type RunningPg } from "../src/postgres-container.ts";
+
+const META = {
+  "metadata.root": {
+    package: "acme::studio",
+    children: [
+      { "object.entity": { name: "Project", children: [
+        { "source.rdb": { "@table": "projects" } },
+        { "field.long": { name: "id" } },
+        { "field.string": { name: "name" } },
+        { "identity.primary": { name: "id", "@fields": "id" } },
+      ] } },
+      { "object.entity": { name: "TimeEntry", children: [
+        { "source.rdb": { "@table": "time_entries" } },
+        { "field.long": { name: "id" } },
+        { "field.long": { name: "projectId" } },
+        { "field.decimal": { name: "hours", "@precision": 6, "@scale": 2 } },
+        { "identity.primary": { name: "id", "@fields": "id" } },
+        { "identity.reference": { name: "project", "@fields": "projectId", "@references": "Project" } },
+      ] } },
+      { "object.projection": { name: "ProjectHours", children: [
+        { "source.rdb": { "@kind": "view", "@view": "v_project_hours" } },
+        { "field.long": { name: "id", extends: "Project.id" } },
+        { "field.string": { name: "name", extends: "Project.name" } },
+        { "field.decimal": { name: "totalHours", "@precision": 10, "@scale": 2, children: [
+          { "origin.aggregate": { "@agg": "sum", "@of": "TimeEntry.hours", "@via": "Project.timeEntries" } } ] } },
+        { "field.int": { name: "entryCount", children: [
+          { "origin.aggregate": { "@agg": "count", "@of": "TimeEntry.id", "@via": "Project.timeEntries" } } ] } },
+        { "field.decimal": { name: "latestHours", "@precision": 6, "@scale": 2, children: [
+          { "origin.first": { "@of": "TimeEntry.hours", "@via": "Project.timeEntries", "@orderBy": ["id:desc"] } } ] } },
+        { "identity.primary": { name: "id", extends: "Project.id" } },
+      ] } },
+    ],
+  },
+};
+
+let pg: RunningPg;
+let k: Kysely<Record<string, unknown>>;
+let root: MetaRoot;
+
+beforeAll(async () => {
+  pg = await startPostgres();
+  k = new Kysely<Record<string, unknown>>({
+    dialect: new PostgresDialect({ pool: new Pool({ connectionString: pg.connectionUri, max: 1 }) }),
+  });
+  const loaded = await new MetaDataLoader().load([new InMemoryStringSource(JSON.stringify(META))]);
+  expect(loaded.errors).toEqual([]);
+  root = loaded.root;
+}, 120_000);
+
+afterAll(async () => {
+  await k.destroy();
+  await pg.stop();
+}, 60_000);
+
+describe("origin @via over a reverse identity.reference hop (postgres)", () => {
+  test("the view carries the aggregate and first columns, joined from the holder", async () => {
+    const views = buildProjectionViews(root, { dialect: "postgres" });
+    const result = await diff({
+      expected: buildExpectedSchema(root, { views }),
+      actual: await introspectPostgres(k),
+      dialect: "postgres",
+      unmanagedNames: collectUnmanagedNames(root),
+    });
+    expect(result.blocked).toEqual([]);
+    for (const stmt of splitSqlStatements(emit(result.changes, { dialect: "postgres" }).up)) {
+      await sql.raw(stmt).execute(k);
+    }
+
+    await sql.raw(`INSERT INTO projects (id, name) VALUES (1, 'Alpha'), (2, 'Idle')`).execute(k);
+    await sql.raw(`INSERT INTO time_entries (id, project_id, hours) VALUES (10, 1, 2.50), (11, 1, 4.00)`).execute(k);
+
+    const rows = await sql<Record<string, unknown>>`SELECT * FROM v_project_hours ORDER BY id`.execute(k);
+    // Columns follow the default snake_case naming; pg returns numeric and bigint as strings.
+    const num = (v: unknown) => (v === null ? null : Number(v));
+    expect(rows.rows.map((r) => ({
+      id: Number(r.id), name: r.name, totalHours: num(r.total_hours),
+      entryCount: num(r.entry_count), latestHours: num(r.latest_hours),
+    }))).toEqual([
+      { id: 1, name: "Alpha", totalHours: 6.5, entryCount: 2, latestHours: 4 },
+      { id: 2, name: "Idle", totalHours: null, entryCount: 0, latestHours: null },
+    ]);
+  }, 60_000);
+});
diff --git a/server/typescript/packages/metadata/src/loader/validation-passes.ts b/server/typescript/packages/metadata/src/loader/validation-passes.ts
index fe976c190f..ac5b819ef0 100644
--- a/server/typescript/packages/metadata/src/loader/validation-passes.ts
+++ b/server/typescript/packages/metadata/src/loader/validation-passes.ts
@@ -132,6 +132,7 @@ import {
   AGG_ALL,
   AGG_COLLECT,
   ASSEMBLY_ORIGIN_SUBTYPES,
+  COLUMN_VALUE_AGGREGATES,
 } from "../persistence/origin/origin-constants.js";
 import {
   inferExprType,
@@ -1543,6 +1544,21 @@ export function validateOriginPaths(root: MetaData): ParseError[] {
           // an aggregate computes something new (count/sum/…); spec §4 defines
           // agreement for passthrough only.
           const ofTarget = _validateFromPath(of_, root, obj, field.name, src, errors, "origin.aggregate.@of");
+          // A sum/avg/min/max reduces ONE column. A value-object field is several columns
+          // (flattened) or a document (jsonb), so there is no column to reduce, and summing
+          // a Money-shaped value across rows would add amounts in different currencies.
+          if (
+            ofTarget !== undefined &&
+            (COLUMN_VALUE_AGGREGATES as readonly unknown[]).includes(agg) &&
+            // ADR-0039: resolving — @objectRef may be inherited via extends.
+            (ofTarget.field.subType === FIELD_SUBTYPE_OBJECT || ofTarget.field.attr(FIELD_ATTR_OBJECT_REF) !== undefined)
+          ) {
+            errors.push(new ParseError(
+              `origin.aggregate @agg:${String(agg)} on ${obj.name}.${field.name}: @of "${of_}" is a value-object ` +
+                `field (field.${ofTarget.field.subType}), which has no single column to aggregate. Aggregate a ` +
+                `scalar field instead: model the amount as its own field (and group by its currency).`,
+              { code: "ERR_INVALID_ORIGIN", source: resolvedSource(src, `${obj.fqn()}::${field.name}`, of_) }));
+          }
           // #195 — collect preserves the element type: the array field's own subType
           // must equal the @of column's subType (the #185 doctrine on the element).
           if (isCollect && ofTarget !== undefined && field.subType !== ofTarget.field.subType) {
diff --git a/server/typescript/packages/metadata/src/persistence/origin/origin-constants.ts b/server/typescript/packages/metadata/src/persistence/origin/origin-constants.ts
index f2a5ee3aaf..47f994b0d1 100644
--- a/server/typescript/packages/metadata/src/persistence/origin/origin-constants.ts
+++ b/server/typescript/packages/metadata/src/persistence/origin/origin-constants.ts
@@ -91,6 +91,9 @@ export const ORIGIN_FIRST_ATTR_FILTER = "filter";
 export const AGGREGATE_FUNCTIONS = ["count", "sum", "avg", "min", "max"] as const;
 export type AggregateFunction = (typeof AGGREGATE_FUNCTIONS)[number];
 
+/** The @agg values that reduce the VALUES of the @of column (count only counts non-null rows). */
+export const COLUMN_VALUE_AGGREGATES: readonly AggregateFunction[] = ["sum", "avg", "min", "max"];
+
 // #195 — the predicate-quantifier + array-rollup @agg values.
 export const AGG_ANY     = "any";
 export const AGG_ALL     = "all";