Add 10 release, schema, and artifact files

docs/evidence/CHECKLIST_RELEASE_CANDIDATE.md

@@ -0,0 +1,19 @@
+# Release Candidate Checklist Artifact
+
+Artifacts that should exist before a release candidate is declared.
+
+## Checklist
+- [ ] `make full-reproducibility-report` succeeds
+- [ ] `python scripts/check_doc_links.py` — 0 broken links
+- [ ] `python scripts/safety/check_claims.py` — 0 overclaiming
+- [ ] `python3 -m pytest -q` — all tests pass
+- [ ] PUBLICATION_PACK.md checklist is complete
+- [ ] METRICS_CURRENT.md is up to date
+- [ ] Evidence certificates are regenerated
+- [ ] Lab batch pack is built (`make lab-batch-pack`)
+- [ ] Decision log is current
+- [ ] Limitations overview is accurate
+
+## Related
+- `docs/review/PUBLICATION_PACK.md`
+- `docs/evidence/DOCS_READINESS_CHECKLIST.md`

docs/evidence/GUIDE_MIGRATION_GENERATOR.md

@@ -0,0 +1,30 @@
+# Artifact Migration Guide Generator
+
+Template for creating artifact migration guides.
+
+## Template
+```markdown
+# Migration Guide: [Old Format] → [New Format]
+
+## What Changed
+Brief description of the format change.
+
+## Why
+Why the change was necessary.
+
+## Migration Steps
+1. Step-by-step instructions.
+2. Commands to run.
+3. Verification steps.
+
+## Backward Compatibility
+- Old format supported until: [date or version]
+- Converter available: [path or command]
+
+## Related
+- [Link to schema changelog]
+- [Link to deprecation policy]
+```
+
+## Current Status
+No migrations are currently needed. This template is for future use.

docs/evidence/GUIDE_OUTPUT_CHECKSUM_LOCKFILE.md

@@ -0,0 +1,26 @@
+# Output Checksum Lockfile
+
+Records expected checksums for generated outputs.
+
+## Format
+```text
+sha256:<hash>  outputs/demo_ranked.jsonl
+sha256:<hash>  outputs/demo_report.md
+```
+
+## Usage
+After running a known-good pipeline, record the checksums:
+```bash
+sha256sum outputs/*.jsonl outputs/*.md > outputs/checksums.lockfile
+```
+
+To verify:
+```bash
+sha256sum -c outputs/checksums.lockfile
+```
+
+## Rules
+- Lockfiles should be updated when pipeline behavior intentionally changes.
+- CI should fail if lockfiles don't match (detects unexpected changes).
+- Lockfiles are stored in `outputs/` (git-ignored).
+- If outputs are non-deterministic (timestamps), exclude those files.

docs/evidence/GUIDE_PUBLIC_READINESS_DRY_RUN.md

@@ -0,0 +1,28 @@
+# Public Readiness Dry-Run Command
+
+Check if the project is ready for public sharing.
+
+## Usage
+```bash
+make full-reproducibility-report
+python scripts/check_doc_links.py
+python scripts/safety/check_claims.py
+python3 -m pytest -q
+```
+
+## What It Checks
+- All tests pass
+- No broken doc links
+- No overclaiming language
+- Reproducibility report succeeds
+- Phase 4 artifacts exist
+
+## Exit Codes
+| Code | Meaning |
+|:----:|---------|
+| 0 | Ready for public sharing |
+| Non-zero | Check the specific failure |
+
+## Related
+- `docs/evidence/GUIDE_PUBLIC_READINESS.md`
+- `docs/review/RELEASE_DRY_RUN_GUIDE.md`

docs/evidence/GUIDE_RELEASE_NOTES_GENERATOR.md

@@ -0,0 +1,21 @@
+# Release Notes Generator from Local Metadata
+
+How release notes are generated.
+
+## Current Process
+Release notes are written manually at release time, following the
+format in `docs/evidence/GUIDE_RELEASE_NOTES_WRITING.md`.
+
+## Metadata Sources
+- Git log: `git log --oneline --no-decorate v0.5.N-1..HEAD`
+- GitHub issues: `gh issue list --state closed --milestone "v0.5.N"`
+- GitHub PRs: `gh pr list --state merged --base main`
+
+## Manual Steps
+1. Review git log for notable changes.
+2. Group changes by type (Added, Changed, Fixed).
+3. Write a summary sentence.
+4. Review for overclaiming language.
+
+## Status
+Release notes are currently written manually. Automation is a future goal.

docs/evidence/POLICY_FORMAT_DEPRECATION.md

@@ -0,0 +1,16 @@
+# Artifact Format Deprecation Policy
+
+How artifact formats are deprecated and retired.
+
+## Deprecation Process
+1. Announce the deprecation in the decision log and release notes.
+2. Keep the old format working for one release cycle.
+3. Provide a migration path or converter.
+4. After one release cycle, remove the old format support.
+5. Update all documentation referencing the old format.
+
+## Rules
+- Breaking format changes must be announced at least one release in advance.
+- Deprecated formats must still be readable for one release cycle.
+- Format changes must be documented in the schema changelog.
+- Data in old formats should be migrated automatically when possible.

docs/evidence/REPORT_BACKWARD_COMPAT.md

@@ -0,0 +1,18 @@
+# Backward Compatibility Report
+
+Tracks backward compatibility of pipeline outputs.
+
+| Artifact | Current Version | Breaking Changes Planned | Migration Path |
+|----------|:---------------:|:------------------------:|:---------------|
+| Evidence certificates | 1.0 | None | — |
+| Lab results | 1.0 | None | — |
+| Run manifests | 1.0 | None | — |
+| Batch reports | 1.0 | None | — |
+| Panel CSVs | 1.0 | None | — |
+
+## Rules
+- Breaking changes must be documented in the decision log.
+- A breaking change is any change that would cause downstream code to fail.
+- Adding optional fields is NOT a breaking change.
+- Removing or renaming fields IS a breaking change.
+- Changing field types IS a breaking change.

tests/test_golden_artifact_fixtures.py

@@ -0,0 +1,21 @@
+"""Test golden artifact fixtures match expected schema."""
+import json
+from pathlib import Path
+
+
+def test_evidence_certificate_schema():
+    schema = json.loads(Path("schemas/candidate.schema.json").read_text())
+    assert "candidate_id" in str(schema)
+    assert "scores" in str(schema)
+
+
+def test_lab_result_schema():
+    schema = json.loads(Path("schemas/lab_result.schema.json").read_text())
+    for field in ["candidate_id", "assay_type", "result_value"]:
+        assert field in str(schema), f"Missing {field}"
+
+
+def test_run_manifest_schema():
+    schema = json.loads(Path("schemas/run_manifest.schema.json").read_text())
+    for field in ["run_id", "pipeline_version", "generated_at"]:
+        assert field in str(schema), f"Missing {field}"

tests/test_limitations_coverage.py

@@ -0,0 +1,22 @@
+"""Test that limitations are documented for key areas."""
+from pathlib import Path
+
+
+def test_limitations_overview_exists():
+    assert Path("docs/evidence/LIMITATIONS_OVERVIEW.md").exists()
+
+
+def test_limitations_covers_key_areas():
+    text = Path("docs/evidence/LIMITATIONS_OVERVIEW.md").read_text()
+    for area in ["Model", "Data", "Benchmark", "Calibration", "Safety"]:
+        assert area in text, f"Missing limitations section: {area}"
+
+
+def test_simulation_benchmark_has_limitations():
+    text = Path("docs/evidence/SIMULATION_BENCHMARK.md").read_text()
+    assert "No improvement" in text or "blocked" in text or "experimental" in text
+
+
+def test_metrics_current_has_caveats():
+    text = Path("docs/evidence/METRICS_CURRENT.md").read_text()
+    assert "limitation" in text.lower() or "caveat" in text.lower()

tests/test_schema_changelog.py

@@ -0,0 +1,22 @@
+"""Test schema changelog consistency."""
+import json
+from pathlib import Path
+
+
+def test_schema_files_exist():
+    schema_dir = Path("schemas")
+    schemas = list(schema_dir.glob("*.json"))
+    assert len(schemas) >= 5, f"Only {len(schemas)} schemas found"
+
+
+def test_schemas_are_valid_json():
+    for f in Path("schemas").glob("*.json"):
+        data = json.loads(f.read_text())
+        assert "$schema" in data, f"{f.name} missing $schema"
+        assert "type" in data, f"{f.name} missing type"
+
+
+def test_schemas_have_titles():
+    for f in Path("schemas").glob("*.json"):
+        data = json.loads(f.read_text())
+        assert "title" in data, f"{f.name} missing title"