Clarify Lab README role boundaries

README.ko.md

@@ -37,6 +37,16 @@ InferEdgeOrchestrator operation feed
 - Runtime Intelligence smoke chain: Orchestrator operation context → EdgeEnv telemetry history/regression → AIGuard deterministic evidence → Lab-owned deployment risk report
 - Local Studio: inference validation을 브라우저에서 확인하는 local-first workflow UI
 
+## 역할 경계 한눈에 보기
+
+| 영역 | Lab이 담당하는 일 | Lab이 담당하지 않는 일 |
+|---|---|---|
+| Validation decision | compare/evaluate/report/API/job workflow와 최종 `deployment_decision`을 생성한다. | Forge build, Runtime execution, EdgeEnv registry, AIGuard diagnosis policy, Orchestrator scheduling을 소유하지 않는다. |
+| Forge / Runtime evidence | metadata, manifest, worker response, Runtime result, validation evidence를 ingest하고 정렬한다. | `metadata.json`, `manifest.json`, Runtime `result.json`, compare output contract를 임의 변경하지 않는다. |
+| EdgeEnv regression evidence | comparability, regression, telemetry coverage, replay gap을 deployment risk context로 표시한다. | EdgeEnv registry/comparability ownership을 재계산하거나 comparability-first gate를 우회하지 않는다. |
+| AIGuard / Orchestrator evidence | optional deterministic diagnosis와 supplemental operation context를 Lab-owned report에 보존한다. | AIGuard/Orchestrator를 final decision owner로 만들거나 production remote execution/control-plane readiness를 주장하지 않는다. |
+| Local Studio / API | local-first evidence replay, in-memory job, reviewer workflow UI를 제공한다. | production SaaS, DB/queue/auth/billing/upload service, cloud dashboard가 되지 않는다. |
+
 ## What Makes InferEdge Different?
 
 InferEdge는 단순 benchmark tool이 아닙니다.

README.md

@@ -21,6 +21,16 @@ Language: English | [한국어](README.ko.md)
 - Runtime Intelligence smoke chain: Orchestrator operation context -> EdgeEnv telemetry history/regression -> AIGuard deterministic evidence -> Lab-owned deployment risk report
 - Local Studio: interactive workflow UI for inference validation
 
+## Role Boundary At A Glance
+
+| Area | Lab owns | Lab does not own |
+|---|---|---|
+| Validation decision | Compare/evaluate/report/API/job workflow and final `deployment_decision` | Forge build, Runtime execution, EdgeEnv registry, AIGuard diagnosis policy, or Orchestrator scheduling |
+| Forge / Runtime evidence | Ingests and aligns metadata, manifest, worker response, Runtime result, and validation evidence | Mutate `metadata.json`, `manifest.json`, Runtime `result.json`, or compare output contracts |
+| EdgeEnv regression evidence | Renders comparability, regression, telemetry coverage, and replay gaps as deployment risk context | Recompute EdgeEnv registry/comparability ownership or bypass the comparability-first gate |
+| AIGuard / Orchestrator evidence | Preserves optional deterministic diagnosis and supplemental operation context in Lab-owned reports | Make AIGuard/Orchestrator final decision owners or claim production remote execution/control-plane readiness |
+| Local Studio / API | Provides local-first evidence replay, in-memory jobs, and reviewer workflow UI | Become production SaaS, DB/queue/auth/billing/upload service, or cloud dashboard |
+
 ## What Makes InferEdge Different?
 
 InferEdge is not a benchmark tool.

tests/test_ci_doc_guards.py

@@ -45,3 +45,32 @@ def test_validate_changed_files_fails_for_unexpected_tracked_file():
 
     assert "unexpected tracked files" in str(exc_info.value)
     assert "Roadmap.md" in str(exc_info.value)
+
+
+def test_readmes_expose_lab_role_boundaries():
+    root = Path(__file__).resolve().parents[1]
+    readme = (root / "README.md").read_text(encoding="utf-8")
+    readme_ko = (root / "README.ko.md").read_text(encoding="utf-8")
+
+    assert "Language: English | [한국어](README.ko.md)" in readme
+    assert "언어: [English](README.md) | 한국어" in readme_ko
+
+    for required in [
+        "## Role Boundary At A Glance",
+        "final `deployment_decision`",
+        "Mutate `metadata.json`, `manifest.json`, Runtime `result.json`",
+        "bypass the comparability-first gate",
+        "Make AIGuard/Orchestrator final decision owners",
+        "production SaaS, DB/queue/auth/billing/upload service, or cloud dashboard",
+    ]:
+        assert required in readme
+
+    for required in [
+        "## 역할 경계 한눈에 보기",
+        "최종 `deployment_decision`",
+        "`metadata.json`, `manifest.json`, Runtime `result.json`",
+        "comparability-first gate를 우회하지 않는다",
+        "AIGuard/Orchestrator를 final decision owner로 만들거나",
+        "production SaaS, DB/queue/auth/billing/upload service, cloud dashboard",
+    ]:
+        assert required in readme_ko