diff --git a/docs/evidence/GUIDE_ARCHITECTURE_DIAGRAMS.md b/docs/evidence/GUIDE_ARCHITECTURE_DIAGRAMS.md
new file mode 100644
index 00000000..68f2626d
--- /dev/null
+++ b/docs/evidence/GUIDE_ARCHITECTURE_DIAGRAMS.md
@@ -0,0 +1,20 @@
+# Architecture Diagram Source Files
+
+Source files for pipeline architecture diagrams.
+
+| Diagram | Source | Format |
+|---------|--------|--------|
+| Artifact flow | docs/evidence/DIAGRAM_ARTIFACT_FLOW.md | Mermaid |
+| Data flow | docs/evidence/DIAGRAM_DATA_FLOW.md | Mermaid |
+| Dependency graph | docs/evidence/DIAGRAM_DEPENDENCY_GRAPH.md | Mermaid |
+
+## How to Update
+1. Edit the Mermaid diagram in the source file.
+2. Preview the diagram using a Mermaid renderer.
+3. Verify the diagram renders correctly in GitHub (native support).
+
+## Rules
+- All diagrams should use Mermaid format.
+- Keep diagrams focused — no more than 15 nodes.
+- Label all nodes and connections clearly.
+- Update related text when a diagram changes.
diff --git a/docs/evidence/GUIDE_METADATA_CHECKER.md b/docs/evidence/GUIDE_METADATA_CHECKER.md
new file mode 100644
index 00000000..c213e1b5
--- /dev/null
+++ b/docs/evidence/GUIDE_METADATA_CHECKER.md
@@ -0,0 +1,20 @@
+# Package Metadata Completeness Checker
+
+Validates that package metadata is complete.
+
+## What It Checks
+- `name` in pyproject.toml
+- `version` is set
+- `description` is present
+- `requires-python` is specified
+- `license` is declared
+- `dependencies` list exists
+
+## Usage
+```bash
+python scripts/check_pyproject.py
+```
+
+## Related
+- `pyproject.toml` — the project configuration file
+- `docs/evidence/GUIDE_PACKAGE_METADATA.md` — metadata guide
diff --git a/docs/evidence/GUIDE_PYPROJECT_CHECKER.md b/docs/evidence/GUIDE_PYPROJECT_CHECKER.md
new file mode 100644
index 00000000..89ca7aed
--- /dev/null
+++ b/docs/evidence/GUIDE_PYPROJECT_CHECKER.md
@@ -0,0 +1,26 @@
+# Pyproject Consistency Checker
+
+The pyproject consistency checker validates `pyproject.toml` against the actual package.
+
+## Usage
+```bash
+python scripts/check_pyproject.py
+```
+
+## Checks Performed
+- Python version requirement is present
+- Console entrypoints reference existing modules
+- Optional extras reference valid dependencies
+- Package name matches the import name
+- Package can be imported
+
+## Exit Codes
+| Code | Meaning |
+|:----:|---------|
+| 0 | All checks pass |
+| 2 | File not found or parse error |
+| 3 | Inconsistency found |
+
+## Related
+- `scripts/check_pyproject.py` — the checker script
+- `pyproject.toml` — the project configuration file
diff --git a/docs/evidence/GUIDE_RELEASE_DRY_RUN.md b/docs/evidence/GUIDE_RELEASE_DRY_RUN.md
new file mode 100644
index 00000000..23e73610
--- /dev/null
+++ b/docs/evidence/GUIDE_RELEASE_DRY_RUN.md
@@ -0,0 +1,21 @@
+# Final Local Release Dry-Run Script
+
+The release dry-run is performed by:
+
+```bash
+make full-reproducibility-report
+```
+
+This generates `outputs/full_reproducibility_report.json` and `.md`.
+
+## What It Checks
+- Git state (SHA, recent commits)
+- Test collection count
+- Benchmark summaries (500-AMP, cross-dataset, simulations)
+- Simulation module status (all experimental)
+- Phase 4 readiness (8 artifact checks)
+- Honest limitations (6 documented caveats)
+
+## Reading the Report
+See `docs/review/RELEASE_DRY_RUN_GUIDE.md` for a detailed reading guide.
+See `docs/review/RELEASE_DRY_RUN_WALKTHROUGH.md` for an annotated walkthrough.
diff --git a/docs/evidence/GUIDE_REPORT_MANIFEST_MACHINE.md b/docs/evidence/GUIDE_REPORT_MANIFEST_MACHINE.md
new file mode 100644
index 00000000..992f7004
--- /dev/null
+++ b/docs/evidence/GUIDE_REPORT_MANIFEST_MACHINE.md
@@ -0,0 +1,31 @@
+# Machine-Readable Report Manifest
+
+The report manifest is generated by:
+
+```bash
+make full-reproducibility-report
+```
+
+The JSON output at `outputs/full_reproducibility_report.json` includes:
+
+## Top-Level Structure
+```json
+{
+  "report_type": "full_reproducibility",
+  "pipeline_version": "0.1.0",
+  "generated_at": "ISO 8601 timestamp",
+  "loops_completed": 50,
+  "phases_completed": { "...": true },
+  "sections": {
+    "git": { "head_sha": "...", "recent_commits": [...] },
+    "tests": { "collection_output": "..." },
+    "benchmarks": { "...": { "auroc": 0.78 } },
+    "simulation_modules": { "...": "experimental" },
+    "phase_4_readiness": { "artifacts": { "...": "✅" } },
+    "limitations": { "...": "No wet-lab data exists" }
+  }
+}
+```
+
+## Consuming the Report
+The JSON report can be parsed by any JSON-capable tool for automation or dashboards.
diff --git a/docs/evidence/MANIFEST_OPTIONAL_DEPS.md b/docs/evidence/MANIFEST_OPTIONAL_DEPS.md
new file mode 100644
index 00000000..7fcba707
--- /dev/null
+++ b/docs/evidence/MANIFEST_OPTIONAL_DEPS.md
@@ -0,0 +1,26 @@
+# Optional Dependency Manifest
+
+Optional dependencies and their purposes.
+
+| Extra | Package | Purpose |
+|-------|---------|---------|
+| dev | pytest | Test runner |
+| dev | pytest-cov | Test coverage |
+| dev | ruff | Linter |
+| — | PyYAML | Config file parsing |
+| — | jsonschema | Schema validation |
+
+## Installation
+```bash
+# Core only
+pip install openamp-foundry
+
+# With dev tools
+pip install "openamp-foundry[dev]"
+```
+
+## Rules
+- Core dependencies must be kept minimal.
+- Optional dependencies should be documented here.
+- Adding a new optional dependency should be reviewed.
+- No dependency should be optional if it's required for core functionality.
diff --git a/docs/evidence/MATRIX_ADAPTER_CAPABILITY.md b/docs/evidence/MATRIX_ADAPTER_CAPABILITY.md
new file mode 100644
index 00000000..b77a79a9
--- /dev/null
+++ b/docs/evidence/MATRIX_ADAPTER_CAPABILITY.md
@@ -0,0 +1,15 @@
+# Adapter Capability Matrix
+
+Capabilities of external adapters and simulation modules.
+
+| Module | Type | Status | Requires Network | Calibration Data |
+|--------|------|:------:|:----------------:|:----------------:|
+| MembraneProxy | simulation | experimental | No | None |
+| StructureProxy | simulation | experimental | No | None |
+| ExternalSimulationAdapter | adapter | available | Depends | None |
+
+## Rules
+- Experimental modules must not affect candidate ranking.
+- Adapters requiring network must document this.
+- Calibration data must be referenced when available.
+- Modules without calibration data are informational only.
diff --git a/docs/operations/EXTERNAL_VALIDATION_AGENT_PROMPT.md b/docs/operations/EXTERNAL_VALIDATION_AGENT_PROMPT.md
index b3d91c60..60d57f0a 100644
--- a/docs/operations/EXTERNAL_VALIDATION_AGENT_PROMPT.md
+++ b/docs/operations/EXTERNAL_VALIDATION_AGENT_PROMPT.md
@@ -187,4 +187,3 @@ outputs/external_validation/
 consensus matrix joins them all on `candidate_id`, and the report names a ranked
 strict-consensus shortlist with explicit tool-support counts. Report any tool that was
 blocked and why, rather than guessing or fabricating its output.
-```
diff --git a/tests/test_cli_smoke.py b/tests/test_cli_smoke.py
new file mode 100644
index 00000000..4da58420
--- /dev/null
+++ b/tests/test_cli_smoke.py
@@ -0,0 +1,29 @@
+"""Smoke tests for CLI commands."""
+import subprocess
+import sys
+
+
+def test_rank_help():
+    r = subprocess.run([sys.executable, "-m", "openamp_foundry.cli", "rank", "--help"],
+                       capture_output=True, text=True, env={"PYTHONPATH": "src"})
+    assert r.returncode == 0
+    assert "usage" in r.stdout.lower()
+
+
+def test_validate_help():
+    r = subprocess.run([sys.executable, "-m", "openamp_foundry.cli", "validate", "--help"],
+                       capture_output=True, text=True, env={"PYTHONPATH": "src"})
+    assert r.returncode == 0
+    assert "usage" in r.stdout.lower()
+
+
+def test_bench_help():
+    r = subprocess.run([sys.executable, "-m", "openamp_foundry.cli", "bench", "--help"],
+                       capture_output=True, text=True, env={"PYTHONPATH": "src"})
+    assert r.returncode == 0
+
+
+def test_version():
+    from openamp_foundry import __version__
+    assert isinstance(__version__, str)
+    assert len(__version__) > 0
diff --git a/tests/test_format_consistency.py b/tests/test_format_consistency.py
new file mode 100644
index 00000000..aeb9b2c1
--- /dev/null
+++ b/tests/test_format_consistency.py
@@ -0,0 +1,22 @@
+"""Tests for format consistency across generated artifacts."""
+from pathlib import Path
+
+
+def test_jsonl_files_are_valid():
+    """Verify JSONL files in outputs/ have one JSON object per line."""
+    for f in Path("outputs").glob("*.jsonl"):
+        lines = f.read_text().strip().split("\n")
+        for i, line in enumerate(lines, 1):
+            import json
+            try:
+                json.loads(line)
+            except json.JSONDecodeError:
+                assert False, f"Invalid JSON at line {i} in {f}"
+
+
+def test_markdown_files_have_no_broken_code_blocks():
+    """Verify markdown code blocks are properly closed."""
+    for md in Path("docs").rglob("*.md"):
+        text = md.read_text()
+        backtick_count = text.count("```")
+        assert backtick_count % 2 == 0, f"Unclosed code block in {md}"
diff --git a/tests/test_optional_imports.py b/tests/test_optional_imports.py
new file mode 100644
index 00000000..61576165
--- /dev/null
+++ b/tests/test_optional_imports.py
@@ -0,0 +1,18 @@
+"""Test that optional imports fail gracefully when dependencies are missing."""
+
+def test_import_core():
+    """Core package should import without optional dependencies."""
+    import openamp_foundry  # noqa: F401
+
+
+def test_import_cli():
+    from openamp_foundry.cli import main  # noqa: F401
+
+
+def test_import_pipeline():
+    from openamp_foundry.pipeline import score_candidates  # noqa: F401
+
+
+def test_import_simulation():
+    from openamp_foundry.simulation import MembraneProxy  # noqa: F401
+    from openamp_foundry.simulation import StructureProxy  # noqa: F401
diff --git a/tests/test_tutorial_runner.py b/tests/test_tutorial_runner.py
new file mode 100644
index 00000000..3609dc6c
--- /dev/null
+++ b/tests/test_tutorial_runner.py
@@ -0,0 +1,21 @@
+"""Run tutorial commands and verify they work."""
+import subprocess
+import sys
+
+
+def test_demo_help():
+    r = subprocess.run(["make", "demo"], capture_output=True, text=True)
+    # make demo may fail if dependencies aren't installed, but should at least attempt
+    assert r.returncode in (0, 2)
+
+
+def test_quickstart_commands():
+    """Verify key quickstart commands are valid."""
+    import argparse
+    from openamp_foundry.cli.main import build_parser
+    parser = build_parser()
+    # Verify rank subcommand exists
+    for action in parser._actions:
+        if hasattr(action, 'choices') and action.choices:
+            assert 'rank' in action.choices
+            break