test(opy): consolidate compatibility test metadata

AGENTS.md

@@ -13,11 +13,11 @@ or LPP, while `opy-rs` must remain independently usable as a library and CLI.
 `opy-rs` owns:
 
 - OverPy syntax, lexer, parser, source model, preprocessing, macros,
-  semantic resolution, diagnostics, provenance, and OPY HIR;
+  semantic resolution, diagnostics, source mapping, and OPY HIR;
 - OverPy-specific compiler/lowering semantics and backend-affecting behavior;
 - standalone OPY tooling APIs and CLI surfaces;
 - Workshop → OPY reconstruction when implemented;
-- OPY compatibility evidence and support claims.
+- OPY compatibility tests and support claims.
 
 `workshop-rs` owns:
 
@@ -38,7 +38,7 @@ into this repository. Never add a dependency from `workshop-rs` back to
 The standalone source-analysis path must remain usable without requiring
 Workshop emission. Compilation may depend on `workshop-rs`; `check`, semantic
 inspection, source queries, and other Workshop-independent operations should
-not be forced through the compiler pipeline without an evidence-backed need.
+not be forced through the compiler pipeline without a concrete requirement.
 
 Do not invent WrightKit-only OPY syntax. Compatibility is observable semantics,
 not output-text identity, optimizer implementation, formatting, temporary
@@ -63,7 +63,7 @@ If the Issue, current contract, and source/tests disagree materially, stop and
 surface the mismatch rather than selecting a design by implementation
 convenience.
 
-## Upstream reference and provenance
+## Upstream reference and source attribution
 
 For the declared OverPy core-language surface, the established upstream OverPy
 implementation is the executable specification. Core behavior is presumptively
@@ -72,11 +72,11 @@ or demonstrated to be a non-contractual implementation artifact.
 
 Inspect upstream implementation, docs, and tests to understand behavior; then
 implement the behavior directly in clear Rust and verify it through
-compatibility evidence. Do not mechanically translate, import, link, or bundle
+compatibility tests and reference comparisons. Do not mechanically translate, import, link, or bundle
 upstream implementation/data into the `opy-rs` core or release artifacts.
 
 OverPy is GPL-3.0; this repository is AGPL-3.0-or-later. Pinned identity,
-fixture/probe provenance, and licensing boundaries are documented in
+fixture/probe attribution, and licensing boundaries are documented in
 [`docs/compatibility/upstream-references.md`](docs/compatibility/upstream-references.md).
 
 Inventories, manifests, probes, differential tests, corpus fixtures, and real
@@ -89,7 +89,7 @@ Prefer typed Rust for observable OverPy behavior and invariants: receiver/member
 semantics, argument binding, contextual dispatch, macro/directive behavior,
 coercion/evaluation rules, and special lowering.
 
-Declarative data may carry large mechanical inventories, names, provenance, and
+Declarative data may carry large mechanical inventories, names, source attribution, and
 facts that do not themselves program language behavior. Do not extend an
 existing manifest/registry with new semantic control fields merely because the
 current implementation already contains similar metadata; existing
@@ -104,7 +104,7 @@ project exposes a blocker:
 2. fix it here if the missing behavior is OverPy-owned;
 3. route missing canonical Workshop behavior to `workshop-rs`;
 4. preserve a minimized regression where practical while keeping the full
-   project evidence;
+   project test coverage;
 5. do not split implementation work into smaller issues solely for bookkeeping
    when one coherent change can be reviewed and validated safely.
 
@@ -122,8 +122,8 @@ cargo test --workspace --all-targets --all-features
 python3 -m unittest discover -s tools/overpy/tests
 ```
 
-Oracle-required compatibility probes run separately against the pinned
-reference. A local test count is not sufficient evidence when a change claims
+Oracle-required compatibility tests run separately against the pinned
+reference. A local test count is not sufficient when a change claims
 real-project compatibility; rerun the affected full-project workflow.
 
 ## Delivery
@@ -132,6 +132,6 @@ real-project compatibility; rerun the affected full-project workflow.
   through PRs.
 - Keep commits focused and issue-linked where an issue exists.
 - Keep compatibility/support claims synchronized with actual implementation and
-  evidence.
+  tests.
 - Never commit credentials, private runtime data, or unreviewed third-party
   material.

README.md

@@ -9,7 +9,7 @@ or as a Language Provider Protocol (LPP) process for extended linting, analysis,
 and editor support.
 
 `opy-rs` owns OverPy syntax, preprocessing, macros, semantic resolution,
-compiler lowering, diagnostics, provenance, and source reconstruction. Shared
+compiler lowering, diagnostics, source mapping, and source reconstruction. Shared
 Workshop semantics, catalog identities, and emission remain delegated to
 `workshop-rs`.
 
@@ -84,7 +84,7 @@ or upstream internal architecture. Optimizer behavior that materially changes
 Workshop structure or element cost remains in scope; only incidental optimizer
 shape is outside the target. Support claims are backed by the
 canonical [language-support contract](docs/language-support.md), the corpus,
-and pinned OverPy reference evidence.
+and pinned OverPy reference results.
 
 > [!IMPORTANT]
 > `opy-rs` follows the OverPy language. It does not introduce a WrightKit-only
@@ -118,7 +118,7 @@ Python test suites.
 
 ## Documentation
 
-Current architecture, compatibility evidence, APIs, HIR, provenance, and
+Current architecture, compatibility tests, APIs, HIR, source mapping, and
 maintainer references are indexed in [`docs/README.md`](docs/README.md).
 
 ## Contributing

crates/opy-cli/src/bin/opy-compat.rs

@@ -1,4 +1,4 @@
-//! Internal compatibility evidence producer.
+//! Internal compatibility comparison producer.
 //!
 //! This target is built only by the compatibility gate. It is deliberately
 //! separate from `opy-cli compile`: pinned oracle input must never become part

crates/opy-rs/src/compiler/tests/resource_baseline.rs

@@ -61,7 +61,7 @@ struct InputIdentity {
     sha256: String,
     source: String,
     parameters: JsonValue,
-    provenance: JsonValue,
+    attribution: JsonValue,
 }
 
 #[derive(Debug)]
@@ -356,7 +356,7 @@ fn synthetic_identity(id: &str, source: String, parameters: JsonValue) -> InputI
         sha256: sha256(&source),
         source: format!("compiler/tests/resource_baseline.rs::{id}"),
         parameters,
-        provenance: serde_json::json!({
+        attribution: serde_json::json!({
             "kind": "original",
             "origin": "WrightKit opy-rs resource audit workload",
             "license": "AGPL-3.0-or-later",
@@ -376,9 +376,9 @@ fn real_world_workload() -> Workload {
     .expect("parabola metadata is valid JSON");
     Workload {
         id: "real-world-parabola",
-        description: "provenance-linked OverPy parabola example compiled end to end",
+        description: "attributed OverPy parabola example compiled end to end",
         input: InputIdentity {
-            kind: "provenance-linked-corpus".to_string(),
+            kind: "attributed-corpus".to_string(),
             bytes: source.len(),
             sha256: sha256(&source),
             source: format!("{REAL_FIXTURE}/parabola.opy"),
@@ -389,7 +389,7 @@ fn real_world_workload() -> Workload {
                 ),
                 "entry": "parabola.opy"
             }),
-            provenance: metadata["provenance"].clone(),
+            attribution: metadata["attribution"].clone(),
         },
         run: run_real_world,
     }