Add 10 test, review, and policy files

docs/evidence/GUIDE_PR_SAFETY_CHECK.md

@@ -0,0 +1,21 @@
+# PR Template Check for Safety-Sensitive Files
+
+PRs that modify safety-sensitive files require additional checks.
+
+## Safety-Sensitive Files
+- `SAFETY.md`
+- `RESPONSIBLE_USE.md`
+- `docs/trust/*`
+- `AGENTS.md`
+- `configs/recalibration_policy.yaml`
+- `configs/pipeline.yaml`
+
+## Required Checks
+- [ ] Change has been reviewed by a safety reviewer.
+- [ ] No overclaiming language introduced (run `check_claims.py`).
+- [ ] Safety implications documented in PR description.
+- [ ] All existing safety tests pass.
+
+## Enforcement
+- PR template should include a section for safety-sensitive changes.
+- CI should flag PRs that touch safety-sensitive files without a safety reviewer.

docs/evidence/GUIDE_PUBLIC_SUMMARY_GENERATOR.md

@@ -0,0 +1,34 @@
+# Public Summary Generator for Review Packets
+
+Generates a public summary from review packet contents.
+
+## Inputs
+- Evidence certificates for selected candidates
+- LIMITATIONS_OVERVIEW.md
+- METRICS_CURRENT.md (summary)
+- The specific claim being reviewed
+
+## Output Format
+```markdown
+# Summary: [Claim]
+
+## What Was Done
+One paragraph description.
+
+## Key Results
+- Metric: value
+- Key finding: statement
+
+## Limitations
+- Known caveat 1
+- Known caveat 2
+
+## Next Steps
+- What happens next.
+```
+
+## Rules
+- The summary must not overclaim.
+- Run `check_claims.py` on the generated summary.
+- Limitations should be as prominent as results.
+- The summary should be understandable by a non-specialist.

docs/evidence/GUIDE_REVIEW_PACKET_DIFF.md

@@ -0,0 +1,22 @@
+# Review Packet Diff Report
+
+Reports differences between review packet versions.
+
+## Format
+```diff
++ Added: SEED-020_VAR_004 (new candidate)
+- Removed: SEED-015_VAR_002 (deprioritized)
+~ Changed: LIMITATIONS_OVERVIEW.md (updated simulation section)
+```
+
+## When to Generate
+- When a review packet is updated after initial review.
+- When new candidates are added to the panel.
+- When benchmark results change significantly.
+- Before sending the final packet to reviewers.
+
+## Rules
+- Changes should be highlighted, not hidden.
+- Reviewers should be notified of significant changes.
+- The diff should be included in the review packet cover note.
+- If changes are extensive, consider a new review cycle.

docs/evidence/POLICY_SLOW_TEST.md

@@ -0,0 +1,25 @@
+# Slow-Test Marker Policy
+
+How to mark and handle slow tests.
+
+## Marking
+```python
+@pytest.mark.slow
+def test_expensive_operation():
+    ...
+```
+
+## Running
+```bash
+# Skip slow tests
+pytest -m "not slow"
+
+# Only slow tests
+pytest -m slow
+```
+
+## Rules
+- Tests that take > 5s should be marked as slow.
+- CI should run both slow and fast tests.
+- Local development should default to skipping slow tests.
+- Slow tests should be reviewed for optimization opportunities.

docs/evidence/REPORT_TEST_RUNTIME_BUDGET.md

@@ -0,0 +1,20 @@
+# Test Runtime Budget Report
+
+Reports how long tests take to run and flags slow tests.
+
+## Current Status
+Full test suite: ~80 seconds (varies by machine)
+
+## Budget
+| Test Category | Budget | Current |
+|:-------------|:------:|:-------:|
+| Unit tests | < 30s | ✅ |
+| CLI tests | < 30s | ✅ |
+| Integration tests | < 30s | ✅ |
+| Full suite | < 120s | ✅ |
+
+## Rules
+- If a test consistently takes > 5s, mark it as slow.
+- Slow tests should be optimized or moved to a separate CI job.
+- Test runtime should be tracked in CI output.
+- If the full suite exceeds 120s, investigate optimization.

docs/evidence/TAXONOMY_REVIEWER_ROLES.md

@@ -0,0 +1,17 @@
+# Reviewer Role Taxonomy
+
+Defines reviewer roles and their responsibilities.
+
+| Role | Can Approve | Required For | Review Depth |
+|------|:-----------:|--------------|:------------:|
+| Contributor | No | All PRs | Comment only |
+| Reviewer | Low-risk | Medium-risk PRs | Functional |
+| Maintainer | High-risk | High-risk PRs | Deep |
+| Safety reviewer | Safety changes | Safety-sensitive files | Full audit |
+| Project lead | All | Final approval | Final |
+
+## Assignment
+- Reviewers should be assigned based on expertise.
+- Safety changes always need a safety reviewer.
+- High-risk changes need at least 2 reviewers.
+- The project lead has final authority on disputes.

docs/evidence/TEMPLATE_ISSUE_BENCHMARK_CHANGE.md

@@ -0,0 +1,31 @@
+# Issue Template for Benchmark Changes
+
+Template for issues that propose benchmark changes.
+
+```markdown
+## Summary
+Brief description of the benchmark change.
+
+## Current Benchmark
+Link to current benchmark card and metrics.
+
+## Proposed Change
+What benchmark is being added, modified, or removed.
+
+## Rationale
+Why this change is needed.
+
+## Metrics
+What metrics will be affected.
+
+## Gate Impact
+Will this change affect the benchmark regression gate?
+
+## Related
+- Link to related PRs or discussions.
+```
+
+## Rules
+- Benchmark changes affecting the gate require maintainer review.
+- New benchmarks must include documented caveats.
+- Benchmark changes should be discussed before implementation.

docs/evidence/TEMPLATE_ISSUE_SCHEMA_CHANGE.md

@@ -0,0 +1,32 @@
+# Issue Template for Artifact Schema Changes
+
+Template for issues that propose schema changes.
+
+```markdown
+## Summary
+Brief description of the proposed schema change.
+
+## Current Schema
+Link to current schema file and relevant fields.
+
+## Proposed Change
+What fields are being added, removed, or changed.
+
+## Rationale
+Why this change is needed.
+
+## Backward Compatibility
+- Is this change backward-compatible?
+- If not, what is the migration plan?
+
+## Affected Artifacts
+List all artifacts that use this schema.
+
+## Related
+- Link to related PRs or discussions.
+```
+
+## Rules
+- Schema changes require maintainer review.
+- Breaking schema changes require a migration plan.
+- New fields should be optional when possible.

tests/test_proof_ladder_validator.py

@@ -0,0 +1,27 @@
+"""Test proof ladder field validation in artifacts."""
+from pathlib import Path
+
+
+def test_evidence_certificate_has_proof_fields():
+    """Evidence certificates should contain fields relevant to proof level."""
+    import json
+    schema = json.loads(Path("schemas/candidate.schema.json").read_text())
+    props = str(schema.get("properties", {}))
+    assert "scores" in props, "Missing scores in evidence schema"
+    assert "candidate" in props or "candidate_id" in props
+
+
+def test_lab_result_has_proof_fields():
+    import json
+    schema = json.loads(Path("schemas/lab_result.schema.json").read_text())
+    required = schema.get("required", [])
+    for field in ["candidate_id", "assay_type", "result_value"]:
+        assert field in required
+
+
+def test_pipeline_version_tracked():
+    """Pipeline version should be tracked in key artifacts."""
+    import json
+    for schema_name in ["candidate.schema.json", "run_manifest.schema.json"]:
+        schema = json.loads(Path(f"schemas/{schema_name}").read_text())
+        assert "version" in str(schema) or "pipeline" in str(schema).lower()

tests/test_release_status_enum.py

@@ -0,0 +1,25 @@
+"""Test release status enum values against registries."""
+from pathlib import Path
+import json
+
+
+def test_exit_code_registry_has_expected_codes():
+    registry = json.loads(Path("configs/exit_code_registry.json").read_text())
+    codes = registry.get("codes", {})
+    for code_str in ["0", "1", "2", "3"]:
+        assert code_str in codes, f"Missing exit code {code_str}"
+        assert "meaning" in codes[code_str]
+
+
+def test_exit_code_meanings_are_consistent():
+    registry = json.loads(Path("configs/exit_code_registry.json").read_text())
+    codes = registry.get("codes", {})
+    meanings = {v["meaning"] for v in codes.values()}
+    assert len(meanings) == len(codes), "Duplicate meanings found"
+
+
+def test_decision_log_enum_values():
+    schema = json.loads(Path("schemas/decision_log.schema.json").read_text())
+    decision_types = schema["properties"]["decision_type"]["enum"]
+    for dt in ["approved", "rejected", "deferred"]:
+        assert dt in schema["properties"]["decision"]["enum"]