feat: upgrade default review model to gpt-5.4 with xhigh reasoning

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>

Author: factory-arman

action.yml

@@ -92,13 +92,13 @@ inputs:
     required: false
     default: "7"
   review_model:
-    description: "Override the model used for code review (e.g., 'claude-sonnet-4-5-20250929', 'gpt-5.1-codex'). Only applies to review flows."
+    description: "Override the model used for code review (e.g., 'claude-sonnet-4-5-20250929', 'gpt-5.4-mini'). Only applies to review flows."
     required: false
-    default: "gpt-5.2"
+    default: "gpt-5.4"
   reasoning_effort:
-    description: "Override reasoning effort for review flows (passed to Droid Exec as --reasoning-effort). If empty and review_model is also empty, the action defaults internally to gpt-5.2 at high reasoning."
+    description: "Override reasoning effort for review flows (passed to Droid Exec as --reasoning-effort). If empty and review_model is also empty, the action defaults internally to gpt-5.4 at xhigh reasoning."
     required: false
-    default: "high"
+    default: "xhigh"
   review_use_validator:
     description: "Enable two-pass review: generate candidate comments to JSON, then validate and post only approved ones."
     required: false

review/action.yml

@@ -15,11 +15,11 @@ inputs:
   review_model:
     description: "Model to use for review"
     required: false
-    default: "gpt-5.2"
+    default: "gpt-5.4"
   reasoning_effort:
     description: "Reasoning effort for review (passed to Droid Exec as --reasoning-effort)"
     required: false
-    default: "high"
+    default: "xhigh"
   output_file:
     description: "Path to write review results JSON"
     required: false

test/modes/tag/review-command.test.ts

@@ -390,7 +390,7 @@ describe("prepareReviewMode", () => {
       (call: unknown[]) => call[0] === "droid_args",
     ) as [string, string] | undefined;
     // When neither REVIEW_MODEL nor REASONING_EFFORT is provided, no --model or --reasoning-effort
-    // flags are added. Defaults are handled by the action.yml inputs (gpt-5.2 / high).
+    // flags are added. Defaults are handled by the action.yml inputs (gpt-5.4 / xhigh).
     expect(droidArgsCall?.[1]).not.toContain("--model");
     expect(droidArgsCall?.[1]).not.toContain("--reasoning-effort");
   });