feat(skill): add codex-code-review for cross-model code review

pipelines/pr-pipeline/SKILL.md

@@ -2,7 +2,7 @@
 name: pr-pipeline
 description: |
   End-to-end pipeline for creating pull requests: Classify Repo, Stage,
-  Review, Commit, Push, Review-Fix Loop (max 3), Create, Verify. For personal repos,
+  Review, Cross-Model Review (Codex), Commit, Push, Review-Fix Loop (max 3), Create, Verify. For personal repos,
   runs iterative /pr-review + fix cycles before PR creation. For protected-org
   repos, human-gated workflow with user confirmation at every step.
   Use when user says "submit PR", "create pull request", "push and PR",
@@ -145,6 +145,26 @@ All findings are auto-fixed. The fix commit is applied to the staged changes bef
 
 **Gate**: Comprehensive review complete. All findings fixed or explained. No unresolved CRITICAL issues.
 
+### Phase 2b: CROSS-MODEL REVIEW (optional)
+
+**Goal**: Get an independent second opinion from OpenAI Codex CLI to catch issues that same-model review might miss.
+
+**Skip condition**: Skip if `codex` CLI is not installed (`which codex` fails), or if user passes `--skip-codex`. This phase is additive -- it never blocks the pipeline, only adds signal.
+
+**Invoke the codex-code-review skill:**
+
+```
+Invoke: /codex-code-review
+Scope: git diff --cached (staged changes)
+```
+
+Codex runs in read-only sandbox mode with GPT-5.4 xhigh reasoning. It produces structured findings (CRITICAL / IMPROVEMENTS / POSITIVE / SUMMARY). Claude assesses each finding before incorporating -- Codex feedback is a second opinion, not authoritative.
+
+**If Codex finds CRITICAL issues that Claude agrees with**: Fix them and re-stage before proceeding.
+**If Codex is unavailable or errors**: Log the skip reason and proceed. This phase must never block the pipeline.
+
+**Gate**: Cross-model review complete (or skipped). Any agreed-upon findings fixed.
+
 ### Phase 3: COMMIT
 
 **Goal**: Create a meaningful commit with conventional format.

skills/INDEX.json

@@ -164,6 +164,25 @@
       "user_invocable": false,
       "version": "2.0.0"
     },
+    "codex-code-review": {
+      "file": "skills/codex-code-review/SKILL.md",
+      "description": "Get a second-opinion code review from OpenAI Codex CLI (GPT-5.4 xhigh reasoning). Invokes codex exec in read-only sandbox mode, structures feedback as CRITICAL / IMPROVEMENTS / POSITIVE / SUMMARY.",
+      "triggers": [
+        "codex review",
+        "second opinion",
+        "code review codex",
+        "gpt review",
+        "cross-model review"
+      ],
+      "category": "code-review",
+      "user_invocable": true,
+      "version": "1.0.0",
+      "pairs_with": [
+        "systematic-code-review",
+        "parallel-code-review",
+        "go-code-review"
+      ]
+    },
     "comment-quality": {
       "file": "skills/comment-quality/SKILL.md",
       "description": "Review and fix comments containing temporal references, development-activity language, or relative comparisons.",