From c02d91d453fbca78896318e5a324080a976974a6 Mon Sep 17 00:00:00 2001
From: OpenCode <opencode@example.com>
Date: Tue, 7 Jul 2026 10:50:34 +0700
Subject: [PATCH] Add synthetic data and toy fixture policy

Closes #457

- docs/evidence/SYNTHETIC_DATA_POLICY.md -- defines toy, synthetic, and real
  data; labeling requirements; allowed/forbidden use; quarantine rules
- 5 tests
---
 docs/evidence/SYNTHETIC_DATA_POLICY.md | 58 ++++++++++++++++++++++++++
 tests/test_synthetic_data_policy.py    | 29 +++++++++++++
 2 files changed, 87 insertions(+)
 create mode 100644 docs/evidence/SYNTHETIC_DATA_POLICY.md
 create mode 100644 tests/test_synthetic_data_policy.py

diff --git a/docs/evidence/SYNTHETIC_DATA_POLICY.md b/docs/evidence/SYNTHETIC_DATA_POLICY.md
new file mode 100644
index 00000000..3fb8906c
--- /dev/null
+++ b/docs/evidence/SYNTHETIC_DATA_POLICY.md
@@ -0,0 +1,58 @@
+# Synthetic Data and Toy Fixture Policy
+
+## Purpose
+
+Synthetic data and toy fixtures are essential for testing, demonstration,
+and workflow development. They must never be mistaken for real biological
+evidence.
+
+## Definitions
+
+| Term | Meaning | Example |
+|------|---------|---------|
+| **Toy fixture** | Minimal dataset shipped with the repo for demo/testing | `examples/sequences/demo_candidates.csv` |
+| **Synthetic data** | Programmatically generated data mimicking real outputs | `examples/lab_results_generator.py` output |
+| **Real data** | Data from qualified wet-lab experiments | — (none yet) |
+
+## Labeling Requirements
+
+Every synthetic or toy data file MUST contain a clear label:
+
+- **CSV/JSONL files**: Include a `disclaimer` or `source` field with `SYNTHETIC` or `demo`.
+- **JSON files**: Include `"disclaimer": "SYNTHETIC — replace with real data"`.
+- **Scripts generating synthetic data**: Output MUST include a disclaimer in every file.
+
+## Allowed Use
+
+- CI/CD testing and regression detection
+- Developer onboarding and demonstration
+- Workflow prototyping and integration testing
+- Benchmark development (when labeled separately from real benchmarks)
+- Documentation examples
+
+## Forbidden Use
+
+- Representing synthetic results as biological evidence
+- Including synthetic data in publication support packages without clear labeling
+- Using toy fixtures to calibrate or tune production scorers
+- Claiming proof-ladder advancement based on synthetic data
+
+## Quarantine
+
+Synthetic data files live in `examples/` or `outputs/` (git-ignored).
+They are never stored in `data/` or `schemas/` to prevent accidental
+inclusion in real-data workflows.
+
+## Relationship to Proof Ladder
+
+Synthetic data sits at **Level 0** (unvalidated). It cannot upgrade any
+proof-ladder claim. A result demonstrated only on synthetic data must be
+described as "demonstrated on synthetic data" with no implication of
+biological validity.
+
+## Related Documents
+
+- `docs/evidence/EVIDENCE_CERTIFICATE.md` — certificate schema (disclaimer field)
+- `docs/evidence/PROOF_LADDER.md` — evidence level definitions
+- `docs/evidence/NEGATIVE_RESULT_ARCHIVE.md` — negative result handling
+- `docs/trust/DATA_GOVERNANCE.md` — data licensing and governance
diff --git a/tests/test_synthetic_data_policy.py b/tests/test_synthetic_data_policy.py
new file mode 100644
index 00000000..572c2cfa
--- /dev/null
+++ b/tests/test_synthetic_data_policy.py
@@ -0,0 +1,29 @@
+"""Tests for synthetic data policy doc."""
+from pathlib import Path
+
+
+def test_file_exists():
+    assert Path("docs/evidence/SYNTHETIC_DATA_POLICY.md").exists()
+
+
+def test_has_definitions():
+    text = Path("docs/evidence/SYNTHETIC_DATA_POLICY.md").read_text()
+    assert "Toy fixture" in text
+    assert "Synthetic data" in text
+    assert "Real data" in text
+
+
+def test_has_allowed_use():
+    text = Path("docs/evidence/SYNTHETIC_DATA_POLICY.md").read_text()
+    assert "Allowed Use" in text
+
+
+def test_has_forbidden_use():
+    text = Path("docs/evidence/SYNTHETIC_DATA_POLICY.md").read_text()
+    assert "Forbidden Use" in text
+
+
+def test_disclaimer_required():
+    text = Path("docs/evidence/SYNTHETIC_DATA_POLICY.md").read_text()
+    assert "SYNTHETIC" in text
+    assert "disclaimer" in text