diff --git a/extensions/qa-lab/src/scenario-catalog-compaction.test.ts b/extensions/qa-lab/src/scenario-catalog-compaction.test.ts
index e6a8d5cde1c5..3ba872955188 100644
--- a/extensions/qa-lab/src/scenario-catalog-compaction.test.ts
+++ b/extensions/qa-lab/src/scenario-catalog-compaction.test.ts
@@ -27,7 +27,7 @@ describe("qa compaction scenario catalog", () => {
     expect(scenario.gatewayConfigPatch).toMatchObject({
       agents: { defaults: { compaction: { mode: "default" } } },
     });
-    expect(flow).toContain("OPENCLAW_QA_FORCE_RUNTIME === 'openclaw'");
+    expect(flow).toContain("env.runtimeId === 'openclaw'");
     expect(flow).toContain("initialRequests[0].errorCode === 'context_length_exceeded'");
     expect(flow).toContain("initialRequests.length === 2");
     expect(flow).toContain("compactionSummaryRequests.length === 2");
@@ -85,7 +85,7 @@ describe("qa compaction scenario catalog", () => {
       | undefined;
     expect(firstAction).toBeDefined();
     const conditional = firstAction?.if as Record<string, unknown> | undefined;
-    expect(conditional?.expr).toBe("env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === 'codex'");
+    expect(conditional?.expr).toBe("env.runtimeId === 'codex'");
     expect(conditional?.["then"]).toMatchObject([
       {
         call: "qaImport",
@@ -102,9 +102,7 @@ describe("qa compaction scenario catalog", () => {
       | Record<string, unknown>
       | undefined;
     expect(runtimeGuard?.assert).toMatchObject({
-      expr: expect.stringContaining(
-        "(env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME ?? 'openclaw') === 'openclaw'",
-      ),
+      expr: expect.stringContaining("env.runtimeId === 'openclaw'"),
     });
 
     const knownGapIndex = flow.indexOf(knownGap);
@@ -115,7 +113,7 @@ describe("qa compaction scenario catalog", () => {
     expect(flow).toContain("new qaErrors.QaSuiteScenarioSkipError");
     expect(flow).toContain("seedQaSessionTranscript");
     expect(flow).toContain("sessions.compaction.branch");
-    expect(flow).toContain("OPENCLAW_QA_FORCE_RUNTIME");
+    expect(flow).toContain("env.runtimeId");
     expect(flow).toContain('"transcriptToolName":"write"');
     expect(flow).toContain('"requireSuccessfulTranscriptToolResult":true');
     expect(flow).toContain("outbound.text === config.finalMarker");
diff --git a/extensions/qa-lab/src/scenario-catalog.runtime-pair.test.ts b/extensions/qa-lab/src/scenario-catalog.runtime-pair.test.ts
index e7a2fbb195e3..92335088f271 100644
--- a/extensions/qa-lab/src/scenario-catalog.runtime-pair.test.ts
+++ b/extensions/qa-lab/src/scenario-catalog.runtime-pair.test.ts
@@ -79,7 +79,7 @@ describe("QA runtime-pair scenario catalog", () => {
     const longContextScenario = readQaScenarioById("long-context-progress-watchdog");
     expect(longContextScenario.execution).toMatchObject({ kind: "flow", runtime: "codex" });
     const longContextFlow = JSON.stringify(longContextScenario.execution.flow);
-    expect(longContextFlow).toContain("OPENCLAW_QA_FORCE_RUNTIME");
+    expect(longContextFlow).toContain("env.runtimeId");
     expect(longContextFlow).not.toContain("patchConfig");
   });
 });
diff --git a/extensions/qa-lab/src/suite-run-standard.parity-retry.test.ts b/extensions/qa-lab/src/suite-run-standard.parity-retry.test.ts
index c96408dee5a3..60e488af39d8 100644
--- a/extensions/qa-lab/src/suite-run-standard.parity-retry.test.ts
+++ b/extensions/qa-lab/src/suite-run-standard.parity-retry.test.ts
@@ -131,6 +131,27 @@ beforeEach(() => {
 });
 
 describe("QA runtime parity scenario retry isolation", () => {
+  it.each([
+    { forcedRuntime: undefined, expectedRuntime: "openclaw" },
+    { forcedRuntime: "codex" as const, expectedRuntime: "codex" },
+  ])(
+    "records $expectedRuntime as the selected runtime fact",
+    async ({ forcedRuntime, expectedRuntime }) => {
+      const runScenario = vi.fn<QaSuiteScenarioRunner>().mockImplementation(async (env) => {
+        expect(env.runtimeId).toBe(expectedRuntime);
+        return makeRetryTestResult("pass");
+      });
+
+      await runQaFlowSuiteStandard(
+        { lab: makeRetryTestLab(), ...(forcedRuntime ? { forcedRuntime } : {}) },
+        makeRetryTestContext(),
+        runScenario,
+      );
+
+      expect(runScenario).toHaveBeenCalledOnce();
+    },
+  );
+
   it("skips connected-transport readiness for intentionally unhealthy startup", async () => {
     const context = makeRetryTestContext();
     context.gatewayRuntimeOptions = { allowUnhealthyStartup: true };
diff --git a/extensions/qa-lab/src/suite-run-standard.ts b/extensions/qa-lab/src/suite-run-standard.ts
index e1afad5595e5..3480b0a1e794 100644
--- a/extensions/qa-lab/src/suite-run-standard.ts
+++ b/extensions/qa-lab/src/suite-run-standard.ts
@@ -152,6 +152,7 @@ export async function runQaFlowSuiteStandard(
       lab,
       mock: activeMock,
       gateway: activeGateway,
+      runtimeId: params?.forcedRuntime ?? "openclaw",
       outputDir,
       // YAML scenarios should see the full staged gateway config, not just
       // the transport fragment. Routing/session/plugin assertions depend on it.
diff --git a/extensions/qa-lab/src/suite-types.ts b/extensions/qa-lab/src/suite-types.ts
index c0627dea653e..58cdeeb0cc9a 100644
--- a/extensions/qa-lab/src/suite-types.ts
+++ b/extensions/qa-lab/src/suite-types.ts
@@ -26,6 +26,7 @@ export type QaSuiteScenarioResult = {
 
 export type QaSuiteEnvironment = {
   lab: QaLabServerHandle;
+  runtimeId: RuntimeId;
   webSessionIds: Set<string>;
 } & QaSuiteRuntimeEnv;
 
diff --git a/extensions/qa-lab/src/suite.ts b/extensions/qa-lab/src/suite.ts
index 575c67662bef..af5f4fb2d619 100644
--- a/extensions/qa-lab/src/suite.ts
+++ b/extensions/qa-lab/src/suite.ts
@@ -74,6 +74,7 @@ export type QaSuiteScenarioResult = {
 
 type QaSuiteEnvironment = {
   lab: QaLabServerHandle;
+  runtimeId: RuntimeId;
   webSessionIds: Set<string>;
 } & QaSuiteRuntimeEnv;
 
diff --git a/qa/scenarios/models/codex-harness-no-meta-leak.yaml b/qa/scenarios/models/codex-harness-no-meta-leak.yaml
index 4622251f32cf..e4624e088ae0 100644
--- a/qa/scenarios/models/codex-harness-no-meta-leak.yaml
+++ b/qa/scenarios/models/codex-harness-no-meta-leak.yaml
@@ -65,10 +65,10 @@ flow:
             message:
               expr: "`expected live primary model ${config.requiredModel}, got ${env.primaryModel}`"
         - assert:
-            expr: "env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === config.harnessRuntime"
+            expr: "env.runtimeId === config.harnessRuntime"
             message:
-              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
-      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
+              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.runtimeId}`"
+      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.runtimeId}`"
     - name: keeps codex coordination chatter out of the visible reply
       actions:
         - call: reset
diff --git a/qa/scenarios/runtime/compaction-empty-response-recovery.yaml b/qa/scenarios/runtime/compaction-empty-response-recovery.yaml
index 4e05a3e35a56..e19189ea085e 100644
--- a/qa/scenarios/runtime/compaction-empty-response-recovery.yaml
+++ b/qa/scenarios/runtime/compaction-empty-response-recovery.yaml
@@ -39,7 +39,7 @@ flow:
     - name: retries an empty compaction summary at the compaction owner
       actions:
         - assert:
-            expr: "env.providerMode === 'mock-openai' && env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === 'openclaw' && Boolean(env.mock)"
+            expr: "env.providerMode === 'mock-openai' && env.runtimeId === 'openclaw' && Boolean(env.mock)"
             message: compaction output recovery requires the mock-openai OpenClaw runtime
         - call: waitForGatewayHealthy
           args:
diff --git a/qa/scenarios/runtime/compaction-reasoning-only-recovery.yaml b/qa/scenarios/runtime/compaction-reasoning-only-recovery.yaml
index 16d38baf8b1f..dd2dc9b059e2 100644
--- a/qa/scenarios/runtime/compaction-reasoning-only-recovery.yaml
+++ b/qa/scenarios/runtime/compaction-reasoning-only-recovery.yaml
@@ -39,7 +39,7 @@ flow:
     - name: retries a reasoning-only compaction summary at the compaction owner
       actions:
         - assert:
-            expr: "env.providerMode === 'mock-openai' && env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === 'openclaw' && Boolean(env.mock)"
+            expr: "env.providerMode === 'mock-openai' && env.runtimeId === 'openclaw' && Boolean(env.mock)"
             message: compaction output recovery requires the mock-openai OpenClaw runtime
         - call: waitForGatewayHealthy
           args:
diff --git a/qa/scenarios/runtime/compaction-retry-mutating-tool.yaml b/qa/scenarios/runtime/compaction-retry-mutating-tool.yaml
index c2e91a2e000e..d1de13188cc8 100644
--- a/qa/scenarios/runtime/compaction-retry-mutating-tool.yaml
+++ b/qa/scenarios/runtime/compaction-retry-mutating-tool.yaml
@@ -48,7 +48,7 @@ flow:
     - name: compacts bulky history and mutates exactly once after overflow
       actions:
         - if:
-            expr: "env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === 'codex'"
+            expr: "env.runtimeId === 'codex'"
             then:
               - call: qaImport
                 args:
@@ -57,7 +57,7 @@ flow:
               - throw:
                   expr: "new qaErrors.QaSuiteScenarioSkipError('known-harness-gap compaction-retry-mutating-tool: provider-error recovery does not invoke Codex native compaction; native token-threshold compaction needs a separate scenario.')"
         - assert:
-            expr: "(env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME ?? 'openclaw') === 'openclaw' && env.providerMode === 'mock-openai' && Boolean(env.mock)"
+            expr: "env.runtimeId === 'openclaw' && env.providerMode === 'mock-openai' && Boolean(env.mock)"
             message: compaction overflow injection requires the mock-openai OpenClaw runtime
         - call: waitForGatewayHealthy
           args:
diff --git a/qa/scenarios/runtime/long-context-progress-watchdog.yaml b/qa/scenarios/runtime/long-context-progress-watchdog.yaml
index 7d2d60357a0e..b494c7ecdfbe 100644
--- a/qa/scenarios/runtime/long-context-progress-watchdog.yaml
+++ b/qa/scenarios/runtime/long-context-progress-watchdog.yaml
@@ -44,9 +44,9 @@ flow:
             - ref: env
             - 60000
         - assert:
-            expr: "env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === config.harnessRuntime"
+            expr: "env.runtimeId === config.harnessRuntime"
             message:
-              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
+              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.runtimeId}`"
         - call: reset
         - set: logCursor
           value:
diff --git a/qa/scenarios/workspace/medium-game-plan-codex-harness.yaml b/qa/scenarios/workspace/medium-game-plan-codex-harness.yaml
index 448591863bd1..60c642241753 100644
--- a/qa/scenarios/workspace/medium-game-plan-codex-harness.yaml
+++ b/qa/scenarios/workspace/medium-game-plan-codex-harness.yaml
@@ -68,10 +68,10 @@ flow:
             message:
               expr: "`expected live primary model ${config.requiredModel}, got ${env.primaryModel}`"
         - assert:
-            expr: "env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === config.harnessRuntime"
+            expr: "env.runtimeId === config.harnessRuntime"
             message:
-              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
-      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
+              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.runtimeId}`"
+      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.runtimeId}`"
     - name: builds the medium game artifact
       actions:
         - call: reset
diff --git a/qa/scenarios/workspace/medium-game-plan-openclaw-harness.yaml b/qa/scenarios/workspace/medium-game-plan-openclaw-harness.yaml
index 609cf28e656c..c916b423e5b6 100644
--- a/qa/scenarios/workspace/medium-game-plan-openclaw-harness.yaml
+++ b/qa/scenarios/workspace/medium-game-plan-openclaw-harness.yaml
@@ -66,10 +66,10 @@ flow:
             message:
               expr: "`expected live primary model ${config.requiredModel}, got ${env.primaryModel}`"
         - assert:
-            expr: "env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME === config.harnessRuntime"
+            expr: "env.runtimeId === config.harnessRuntime"
             message:
-              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
-      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.gateway.runtimeEnv.OPENCLAW_QA_FORCE_RUNTIME}`"
+              expr: "`expected suite runtime ${config.harnessRuntime}, got ${env.runtimeId}`"
+      detailsExpr: "`provider=${selected?.provider} model=${selected?.model} runtime=${env.runtimeId}`"
     - name: builds the medium game artifact
       actions:
         - call: reset