Add release dry-run reading guide

docs/review/RELEASE_DRY_RUN_GUIDE.md

@@ -0,0 +1,84 @@
+# Release Dry-Run Reading Guide
+
+## Purpose
+
+The release dry-run (`make full-reproducibility-report`) generates a report
+that summarizes pipeline state, benchmark results, simulation module status,
+Phase 4 readiness, and honest limitations — **without publishing or tagging
+anything**. It is a local preview, not a release.
+
+## How to Run
+
+```bash
+make full-reproducibility-report
+```
+
+Output: `outputs/full_reproducibility_report.json` (machine-readable) and
+`outputs/full_reproducibility_report.md` (human-readable).
+
+## Reading the Report
+
+### 1. Git State
+
+Shows the current commit SHA and recent commits. If the SHA doesn't match
+what you expected, you may be on the wrong branch.
+
+### 2. Test Count
+
+Shows the pytest collection count. A drop from the previous release is a
+blocker — investigate before proceeding.
+
+### 3. Benchmark Summaries
+
+| Section | What to look for |
+|---------|------------------|
+| `benchmark_500` | AUROC should be consistent with METRICS_CURRENT.md |
+| `cross_dataset` | DRAMP AUROC should be within 0.03 of primary |
+| `simulation_ablation` | Verdict should be NO_IMPROVEMENT (expected) |
+| `simulation_baselines` | Verdict should be NO_IMPROVEMENT (expected) |
+
+### 4. Simulation Module Status
+
+| Status | Meaning |
+|--------|---------|
+| `experimental` | Module does not beat cheap baselines — expected |
+| `blocked` | Weighted mode correctly prevented by gate |
+| `info available` | Users can inspect scores via `--simulation-mode info` |
+
+### 5. Phase 4 Readiness
+
+Each artifact is checked for existence. A ❌ means the file was not found.
+Common causes:
+
+- **lab_batch_pack**: Run `make lab-batch-pack` to generate the zip.
+- **Other missing files**: Check if the path is correct in the report script.
+
+If any artifact shows ❌ and it should exist, fix the path or regenerate.
+
+### 6. Honest Limitations
+
+This section lists known caveats. It is not a blocker — it is a truthfulness
+check. Every limitation should be accurate and current.
+
+## Interpreting the Verdict
+
+| Status | Meaning |
+|--------|---------|
+| ✅ All checks pass | Pipeline is ready for review or release consideration |
+| ⚠️ Warnings | Non-blocking issues (check each warning) |
+| ❌ Blocker | Must be resolved before any public claim or release |
+
+## What the Dry-Run Does NOT Do
+
+- It does not publish anything.
+- It does not tag or create a release.
+- It does not send data anywhere.
+- It does not change any files outside `outputs/`.
+
+## Next Actions After a Clean Report
+
+1. Review the honest limitations section for accuracy.
+2. Run `make lab-batch-pack` if the zip is missing.
+3. Verify benchmark numbers against `docs/evidence/METRICS_CURRENT.md`.
+4. Proceed to human review or publication preparation per
+   `docs/review/PUBLICATION_PACK.md`.

tests/test_release_dry_run_guide.py

@@ -0,0 +1,26 @@
+"""Tests for release dry-run reading guide."""
+from pathlib import Path
+
+
+def test_file_exists():
+    assert Path("docs/review/RELEASE_DRY_RUN_GUIDE.md").exists()
+
+
+def test_has_run_command():
+    text = Path("docs/review/RELEASE_DRY_RUN_GUIDE.md").read_text()
+    assert "make full-reproducibility-report" in text
+
+
+def test_has_verdict_section():
+    text = Path("docs/review/RELEASE_DRY_RUN_GUIDE.md").read_text()
+    assert "Interpreting the Verdict" in text
+
+
+def test_has_does_not_do():
+    text = Path("docs/review/RELEASE_DRY_RUN_GUIDE.md").read_text()
+    assert "does not publish" in text
+
+
+def test_has_next_actions():
+    text = Path("docs/review/RELEASE_DRY_RUN_GUIDE.md").read_text()
+    assert "Next Actions" in text