cschanhniem · Jul 9, 2026 · Jul 9, 2026
diff --git a/Makefile b/Makefile
@@ -1,4 +1,4 @@
-.PHONY: help demo test lint ci clean bench-leakage bench-multi-negatives bench-baseline bench-hidden-active bench-cluster-split bench-expert-ablation bench-expert-ablation-500 bench-selectivity bench-feature-decomp bench-gate bench-easy-baseline bench-charge-matched bench-order-dependent bench-precision-at-k bench-per-family bench-simulation-ablation bench-simulation-ablation-within-amp simulation-gate regenerate-all generate phase3 pilot validate-scoring validate-scoring-phase3 validate-scoring-strict external-predict pilot-confident presynth-qc gold-standard diversity synthesis-order novelty-broad external-consensus questionnaire gate-check ip-report benchmark-card wave0-5-gate-check wave0-5-novelty-audit wave0-5-novelty-audit-v2 wave0-5-panel wave0-5-evidence wave0-5-fill-external wave0-5b-generate wave0-5b-filter recalibration-engine recalibration-engine-dry-run validate-policy-version generate-synthetic-lab-results bump-policy-version
+.PHONY: help demo test lint ci clean bench-leakage bench-multi-negatives bench-baseline bench-hidden-active bench-cluster-split bench-expert-ablation bench-expert-ablation-500 bench-selectivity bench-feature-decomp bench-gate bench-easy-baseline bench-charge-matched bench-order-dependent bench-precision-at-k bench-per-family bench-simulation-ablation bench-simulation-ablation-within-amp simulation-gate regenerate-all generate phase3 pilot validate-scoring validate-scoring-phase3 validate-scoring-strict external-pilot pilot-confident presynth-qc gold-standard diversity synthesis-order novelty-broad external-consensus questionnaire gate-check ip-report benchmark-card wave0-5-gate-check wave0-5-novelty-audit wave0-5-novelty-audit-v2 wave0-5-panel wave0-5-evidence wave0-5-fill-external wave0-5b-generate wave0-5b-filter recalibration-engine recalibration-engine-dry-run validate-policy-version generate-synthetic-lab-results bump-policy-version generate-review-packet
 
 PYTHON := $(shell [ -f .venv/bin/python ] && echo .venv/bin/python || echo python3)
 PYTEST  := $(shell [ -f .venv/bin/pytest ] && echo .venv/bin/pytest || echo pytest)
@@ -586,3 +586,12 @@ lab-batch-pack:
 		--evidence-dir outputs/evidence_wave0_5 \
 		--out outputs/lab_batch_pack.zip \
 		--manifest-out outputs/lab_batch_pack_manifest.json
+
+generate-review-packet:
+	PYTHONPATH=src $(PYTHON) scripts/generate_review_packet.py \
+		--pipeline-version v0.5.73 \
+		--git-sha $$(git rev-parse HEAD) \
+		--candidate-count 36 \
+		--proof-ladder-level 2 \
+		--out outputs/review_packet_skeleton.json \
+		--validate
diff --git a/docs/evidence/METRICS_CURRENT.md b/docs/evidence/METRICS_CURRENT.md
@@ -5,7 +5,9 @@ Machine-readable snapshot: `outputs/metrics_snapshot.json` regenerated with `mak
 > **Purpose:** One authoritative table of current pipeline metrics. If any doc disagrees
 > with this file, this file wins. Updated whenever benchmark/benchmark config changes.
 >
-> **Last updated:** 2026-07-09 (Phase E external review schemas — v0.5.72)
+> **Last updated:** 2026-07-09 (Phase E safety release, preregistration, packet generator — v0.5.73)
+> **New in v0.5.73:** Safety-release decision schema (E4), pilot pre-registration schema (E5), review packet generator CLI (E6). 40 tests. 2679 total.
+>
 > **New in v0.5.72:** External review packet schema (E1), example packet (E2), reviewer questionnaire schema (E3). 27 tests. 2666 total.
 >
 > **New in v0.5.71:** Calibration benchmark added — Brier score, reliability diagram, and calibration slope for pipeline ensemble scores. Honest finding: Brier=0.318 (>0.25=uninformative), slope=0.43 (ideal=1.0). Pipeline ranks well but scores are not meaningful probabilities. `make bench-calibration`. 6 tests. ~2100 total.
@@ -54,7 +56,7 @@ Machine-readable snapshot: `outputs/metrics_snapshot.json` regenerated with `mak
 > **New in v0.5.31:** Added dipeptide-order features for sequence-order awareness. `dipeptide_order_score` achieves AUROC 0.7861 on AMP-vs-scrambled discrimination — the strongest order-dependent feature in the pipeline. Only 7/31 features survive scrambling (amphipathicity/helix-wheel + dipeptide). All composition features are purely position-independent (exactly 0.5000 AUROC on scrambled test).
 > **New in v0.5.30:** Easy baseline benchmark added — charge density alone (AUROC 0.8166) outperforms the full pipeline ensemble (0.7792) on AMP-vs-Swiss-Prot-decoy discrimination. Honest finding documented: expected because pipeline optimizes for safety, not raw discrimination.
 > **New in v0.5.29:** Expanded benchmark to 500 AMPs + 500 composition-matched decoys (n=1000). AUROC 0.7792 (CI₉₅: 0.7505–0.8065) confirms signal generalizes. Cluster-aware CI: 0.746–0.8102. Representative AUROC: 0.778. Standard benchmark (n=191) retained for backward comparison.
-> **Pipeline version:** v0.5.72
+> **Pipeline version:** v0.5.73
 > **Branch:** main
 
 ---

diff --git a/docs/research/ROADMAP.md b/docs/research/ROADMAP.md
@@ -1,5 +1,52 @@
 # Roadmap
 
+## v0.5.73 — Loop 72: Phase E E4-E6 — Safety Release, Preregistration, Packet CLI ✓ (2026-07-09)
+
+External review infrastructure: human-gated release decisions, pre-registered pilot
+design, and automated skeleton review packet generation.
+
+Changes:
+- `schemas/safety_release_decision.schema.json` (E4) — JSON Schema for human-review
+  release decisions. Required fields: decision_id, packet_id, decided_at (ISO datetime),
+  decided_by (human name/role), decision (release/hold/reject), rationale (min 20 chars),
+  safety_checks_passed (toxicity/hemolysis/dual_use, all bool), novelty_confirmed (bool),
+  dry_lab_only_acknowledged (must be true), next_steps. Constraint: agent_authored must
+  be false or absent — safety release decisions cannot be made by agents.
+- `schemas/pilot_preregistration.schema.json` (E5) — JSON Schema for freezing pilot
+  experimental design before running assays. Required fields: preregistration_id,
+  packet_id, registered_at, registered_by, assay_types (min 1), pass_criteria (array
+  with metric/threshold/direction, min 1), control_conditions (min 1), sample_size
+  (min 1), blinding, primary_endpoint, secondary_endpoints, statistical_analysis_plan
+  (min 20 chars), amendment_policy, pipeline_version_locked. Prevents HARKing by
+  freezing pass/fail criteria pre-assay — amendment_policy must be explicit.
+- `scripts/generate_review_packet.py` (E6) — CLI script that generates a skeleton
+  external review packet JSON that validates against E1 schema. Flags: --pipeline-version,
+  --git-sha, --candidate-count, --proof-ladder-level, --out, --validate.
+- `Makefile` — Added `generate-review-packet` target.
+- `tests/evidence/test_safety_release_decision_schema.py` — 18 tests covering valid
+  decision, agent_authored=true rejection, missing/empty safety_checks, dry_lab_only
+  enforcement, decision enum, rationale length, decided_by pattern, dissent optional.
+- `tests/evidence/test_pilot_preregistration_schema.py` — 14 tests covering valid
+  preregistration, empty assay_types/pass_criteria/control_conditions rejection,
+  negative sample_size, invalid direction enum, statistical plan length, blinding type.
+- `tests/test_generate_review_packet.py` — 7 tests covering CLI output, schema
+  validation, --validate flag, required fields, invalid args.
+- `tests/evidence/test_negative_result_schema.py` — Updated TestPilotPreregistrationSchema
+  to match new schema fields.
+- `docs/evidence/METRICS_CURRENT.md` — v0.5.73 changelog. Pipeline version bumped.
+  Test count: 2679.
+- `tests/test_test_count_regression.py` — baseline updated to 2679.
+
+Honest boundaries:
+- The safety release schema enforces human-gating on release decisions but does not
+  validate the quality or completeness of the safety review itself.
+- The pilot pre-registration schema prevents post-hoc pass-criteria changes but relies
+  on human discipline to follow the amendment policy.
+- The review packet generator creates a skeleton only — all fields (benchmark results,
+  candidate data, safety attestations) must be updated before use.
+- A schema-valid safety release decision can still be wrong. The schema verifies
+  structure, not correctness.
+
 ## v0.5.72 — Loop 71: External Review Packet Schemas (Phase E, E1-E3) ✓ (2026-07-09)
 
 External credibility infrastructure: machine-checkable schemas that let external

diff --git a/schemas/pilot_preregistration.schema.json b/schemas/pilot_preregistration.schema.json
@@ -1,41 +1,123 @@
 {
   "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "title": "Pilot Pre-Registration",
-  "description": "Schema for pre-registered pilot selection logic. Freezes rules before assay results are known.",
+  "$id": "https://openamp-foundry.org/schemas/pilot_preregistration.schema.json",
+  "title": "OpenAMP Pilot Pre-Registration",
+  "description": "JSON Schema for freezing pilot experimental design before running assays. Pre-registration prevents HARKing (Hypothesizing After Results Known): once registered, pass/fail criteria cannot be changed post-hoc. Amendment policy must be documented explicitly.",
   "type": "object",
   "required": [
-    "pilot_name", "preregistered_at", "preregistered_by",
-    "candidate_pool_source", "selection_rules", "exclusion_criteria",
-    "pass_fail_criteria"
+    "preregistration_id",
+    "packet_id",
+    "registered_at",
+    "registered_by",
+    "assay_types",
+    "pass_criteria",
+    "control_conditions",
+    "sample_size",
+    "blinding",
+    "primary_endpoint",
+    "secondary_endpoints",
+    "statistical_analysis_plan",
+    "amendment_policy",
+    "pipeline_version_locked"
   ],
   "properties": {
-    "pilot_name": {"type": "string", "description": "Human-readable pilot name (e.g., 'Wave 1 Pilot')"},
-    "preregistered_at": {"type": "string", "format": "date-time", "description": "ISO 8601 timestamp"},
-    "preregistered_by": {"type": "string", "description": "Person or agent who created the registration"},
-    "candidate_pool_source": {"type": "string", "description": "Path or description of the candidate pool"},
-    "candidate_pool_hash": {"type": "string", "description": "SHA-256 hash of the pool CSV"},
-    "selection_rules": {
-      "type": "object",
-      "description": "Machine-readable rules used to select candidates"
-    },
-    "exclusion_criteria": {
-      "type": "array",
-      "items": {"type": "string"},
-      "description": "Criteria that would cause a candidate to be excluded"
+    "preregistration_id": {
+      "type": "string",
+      "description": "Unique identifier for this pre-registration (e.g. 'PPR-2026-07-09-001')"
+    },
+    "packet_id": {
+      "type": "string",
+      "description": "Identifier of the external review packet this pre-registration refers to"
+    },
+    "registered_at": {
+      "type": "string",
+      "format": "date-time",
+      "description": "ISO 8601 timestamp when the experimental design was frozen"
     },
-    "pass_fail_criteria": {
-      "type": "object",
-      "description": "Criteria for determining pilot success or failure"
+    "registered_by": {
+      "type": "string",
+      "description": "Name and role of the person who registered this design"
     },
-    "revision_history": {
+    "assay_types": {
       "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "string",
+        "description": "Type of assay to be performed (e.g. 'MIC', 'hemolysis_RBC', 'serum_stability', 'cytotoxicity_mammalian')"
+      },
+      "description": "Types of assays that will be performed. Must have at least one assay type."
+    },
+    "pass_criteria": {
+      "type": "array",
+      "minItems": 1,
       "items": {
         "type": "object",
+        "required": ["metric", "threshold", "direction"],
         "properties": {
-          "date": {"type": "string", "format": "date"},
-          "reason": {"type": "string"}
-        }
-      }
+          "metric": {
+            "type": "string",
+            "description": "The metric against which pass/fail is judged (e.g. 'MIC_50', 'HC50', 'selectivity_index')"
+          },
+          "threshold": {
+            "type": "number",
+            "description": "The numerical threshold value for this criterion"
+          },
+          "direction": {
+            "type": "string",
+            "enum": ["above", "below", "equals"],
+            "description": "Whether passing requires the metric to be above, below, or equal to the threshold"
+          },
+          "notes": {
+            "type": "string",
+            "description": "Optional context or justification for this criterion"
+          }
+        },
+        "description": "A single pass/fail criterion with metric, threshold, and direction"
+      },
+      "description": "Pre-registered pass/fail criteria. These MUST NOT be changed after assays are run. If changes are needed, use the amendment_policy mechanism."
+    },
+    "control_conditions": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "string",
+        "description": "Description of a control condition (e.g. 'Melittin positive control (hemolytic)', 'Vehicle-only negative control')"
+      },
+      "description": "Experimental control conditions that will be used. At least one control is required."
+    },
+    "sample_size": {
+      "type": "integer",
+      "minimum": 1,
+      "description": "Number of biological replicates per condition"
+    },
+    "blinding": {
+      "type": "boolean",
+      "description": "Whether the experimenters will be blinded to candidate identity during assay execution"
+    },
+    "primary_endpoint": {
+      "type": "string",
+      "description": "The primary outcome measure that will determine the study's main conclusion"
+    },
+    "secondary_endpoints": {
+      "type": "array",
+      "items": {
+        "type": "string",
+        "description": "A secondary outcome measure"
+      },
+      "description": "Secondary outcome measures (can be empty if none)"
+    },
+    "statistical_analysis_plan": {
+      "type": "string",
+      "minLength": 20,
+      "description": "Detailed plan for statistical analysis of the results. Must be at least 20 characters."
+    },
+    "amendment_policy": {
+      "type": "string",
+      "description": "Documented policy for how amendments to this pre-registration are handled. Must be explicit about what changes are allowed, who can authorize them, and how they are recorded."
+    },
+    "pipeline_version_locked": {
+      "type": "string",
+      "description": "The exact pipeline version used to generate candidate scores for this pilot (e.g. 'v0.5.72'). Prevents version drift from invalidating selection criteria."
     }
   }
 }
diff --git a/schemas/safety_release_decision.schema.json b/schemas/safety_release_decision.schema.json
@@ -0,0 +1,97 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://openamp-foundry.org/schemas/safety_release_decision.schema.json",
+  "title": "OpenAMP Safety Release Decision",
+  "description": "JSON Schema for the human-review decision on whether to release a candidate panel externally. Safety release decisions CANNOT be made by agents — agent_authored must be absent or false. This ensures release review is auditable and human-gated.",
+  "type": "object",
+  "required": [
+    "decision_id",
+    "packet_id",
+    "decided_at",
+    "decided_by",
+    "decision",
+    "rationale",
+    "safety_checks_passed",
+    "novelty_confirmed",
+    "dry_lab_only_acknowledged",
+    "next_steps"
+  ],
+  "properties": {
+    "decision_id": {
+      "type": "string",
+      "description": "Unique identifier for this release decision (e.g. 'SRD-2026-07-09-001')"
+    },
+    "packet_id": {
+      "type": "string",
+      "description": "Identifier of the external review packet this decision refers to"
+    },
+    "decided_at": {
+      "type": "string",
+      "format": "date-time",
+      "description": "ISO 8601 timestamp when the decision was made"
+    },
+    "decided_by": {
+      "type": "string",
+      "pattern": "^[A-Za-z][A-Za-z .,'()-]+$",
+      "minLength": 3,
+      "description": "Name and/or role of the human who made this decision (e.g. 'Dr. Jane Smith, Safety Lead'). Must be a human, NOT an agent."
+    },
+    "decision": {
+      "type": "string",
+      "enum": ["release", "hold", "reject"],
+      "description": "The release decision: release (approved for external sharing), hold (requires more information), reject (denied release)"
+    },
+    "rationale": {
+      "type": "string",
+      "minLength": 20,
+      "description": "Human-written rationale for the decision. Must be at least 20 characters to ensure meaningful justification."
+    },
+    "safety_checks_passed": {
+      "type": "object",
+      "required": ["toxicity_screened", "hemolysis_screened", "dual_use_reviewed"],
+      "properties": {
+        "toxicity_screened": {
+          "type": "boolean",
+          "description": "Whether candidates were screened for general toxicity signals"
+        },
+        "hemolysis_screened": {
+          "type": "boolean",
+          "description": "Whether hemolysis assay risk was assessed for all candidates"
+        },
+        "dual_use_reviewed": {
+          "type": "boolean",
+          "description": "Whether dual-use implications were reviewed by a qualified human"
+        },
+        "additional_checks": {
+          "type": "object",
+          "description": "Optional additional safety checks beyond the three required ones",
+          "additionalProperties": { "type": "boolean" }
+        }
+      },
+      "description": "Record of which safety checks have been completed and passed"
+    },
+    "novelty_confirmed": {
+      "type": "boolean",
+      "description": "Whether candidate novelty has been confirmed by human review against current literature and databases"
+    },
+    "dry_lab_only_acknowledged": {
+      "type": "boolean",
+      "enum": [true],
+      "description": "REQUIRED: Must be true. Decision acknowledges that all outputs are dry-lab computational hypotheses only and are not biological proof."
+    },
+    "dissent": {
+      "type": "string",
+      "description": "Optional field documenting any reviewer disagreement or dissenting opinion. Required if any reviewer objected to the decision."
+    },
+    "next_steps": {
+      "type": "string",
+      "minLength": 10,
+      "description": "Description of next steps following this decision (e.g., 'Prepare release summary packet', 'Request additional safety data', 'Escalate to external review board')"
+    },
+    "agent_authored": {
+      "type": "boolean",
+      "enum": [false],
+      "description": "SAFETY CONSTRAINT: Must be false or absent. Safety release decisions cannot be made by agents."
+    }
+  }
+}