cschanhniem · Jul 2, 2026 · Jul 2, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -423,6 +423,15 @@ Human review is required before:
 * making external scientific claims;
 * submitting a paper or press release.
 
+## 9. Keep panel-specific evidence separate
+
+Do not collapse external-predictor evidence across panels. The generic Gate 6
+workflow remains pending for any future pilot panel until that panel has filled
+results and a generated report. Wave 0.5 is separate: its completed external
+screen used three submitted activity predictors with CAMPR4 excluded, plus
+separate HemoFinder and AntiCP annotations. When docs disagree, treat
+`docs/METRICS_CURRENT.md` as the current-state source of truth.
+
 ---
 
 # Agent Task Priorities

diff --git a/docs/DECISION_RULES.md b/docs/DECISION_RULES.md
@@ -57,9 +57,14 @@
 |-------|-------|
 | Threshold | ≥3/5 external tools positive |
 | Measure | Manual web submission results |
-| Status | PENDING (see `outputs/external_predict_checklist.md`) |
+| Status | PENDING for any new generic pilot panel until a filled `outputs/external_predict_results.csv` and generated report exist |
 | Action | If FAIL: Expert reviewer override required |
 
+**Scope note:** This generic Gate 6 is not the same as the completed Wave 0.5 external
+screen. Wave 0.5 used three submitted activity predictors with CAMPR4 excluded and
+separate HemoFinder/AntiCP safety annotations; see W0.5-3 and W0.5-4 below plus
+`docs/METRICS_CURRENT.md`.
+
 ## Gate 7 — Human Expert Review
 
 | Field | Value |

diff --git a/docs/EXPERT_REVIEW_PACK.md b/docs/EXPERT_REVIEW_PACK.md
@@ -269,9 +269,13 @@ Before any candidate proceeds to synthesis or assay:
 2. **Pre-registration lock**: PI signs `docs/ASSAY_PREREGISTRATION.md` before ordering synthesis.
 3. **Broad novelty acknowledgement**: File `docs/NOVELTY_BROAD_CHECK.md` with the pre-registration.
    SEED-003 and SEED-001 candidates are near-known AMPs; novelty claims must be appropriately limited.
-4. **External predictor check** (prior to synthesis): Submit `outputs/pilot_panel.fasta` to CAMPR4,
-   AMPScanner v2, dbAMP, AntiCP2, Macrel web server. Fill checklist `outputs/external_predict_checklist.md`.
-   Require ≥3/5 agreement before synthesis order.
+4. **External predictor check** (prior to synthesis): For a new generic pilot panel,
+   submit `outputs/pilot_panel.fasta` to CAMPR4, AMPScanner v2, dbAMP, AntiCP2, and the
+   Macrel web server, then fill `outputs/external_predict_results.csv`. Require ≥3/5
+   agreement before synthesis order. For the Wave 0.5 panel, use the completed
+   current-state summary in `docs/METRICS_CURRENT.md`: CAMPR4 was excluded, so activity
+   consensus is based on three submitted activity predictors plus separate HemoFinder
+   and AntiCP annotations.
 5. **Serum stability first**: Triage all 20 candidates in 50% human serum (0/30/60/120 min) before
    committing to full MIC panel. Pass criterion: >50% intact at 60 min (HPLC/MS primary readout).
 6. **Synthesis order**: After serum stability triage, `make synthesis-order` generates vendor-ready CSV.

diff --git a/docs/EXTERNAL_PREDICTOR_CONSENSUS.md b/docs/EXTERNAL_PREDICTOR_CONSENSUS.md
@@ -26,7 +26,16 @@ classified as **CONFIDENT** and recommended for synthesis.
 
 ## Current Status
 
-**PENDING** — Results require manual web submission for all 5 tools.
+This document describes the **generic pilot-panel consensus workflow**. It remains
+unresolved for any future panel until that panel has a filled
+`outputs/external_predict_results.csv` and a generated consensus report.
+
+Current Wave 0.5 status is different: the completed Wave 0.5 screen is summarized in
+`docs/METRICS_CURRENT.md`. CAMPR4 was not submitted, so Wave 0.5 activity consensus is
+based on three activity predictors (AMPScanner v2, AMPActiPred, Macrel web server) plus
+separate HemoFinder and AntiCP 2.0 safety annotations. Do not treat this generic 5-tool
+template as evidence that Wave 0.5 remains wholly pending.
+
 Macrel v1.6.0 CLI has a known ONNX bug (all candidates classified NAMP — see PR #77);
 use the web server at `big-data-biology.org/software/macrel` instead.
 

diff --git a/docs/REVIEWER_SUMMARY.md b/docs/REVIEWER_SUMMARY.md
@@ -78,7 +78,7 @@ assay is mandatory for all candidates.
 | No 3D structure or docking | Helical assumption may misclassify non-helical mechanisms |
 | Near-seed generation only | Novel sequence space not explored by de novo generation |
 | Melittin safety blind spot | Hemolysis must be assayed — confirmed quantitatively: safety scorer detection AUROC=0.3844 on 14 hemolytic vs 21 selective AMPs (v0.5.9 selectivity benchmark) |
-| No external predictor ensemble yet | CAMPR4/AMPScanner/dbAMP/AntiCP2/Macrel web submission checklist exists at `outputs/external_predict_checklist.md` but results not yet obtained |
+| External predictor evidence is panel-specific | Wave 0.5 external screen is complete for AMPScanner v2, AMPActiPred, Macrel web, HemoFinder, and AntiCP 2.0 with CAMPR4 excluded; a future generic pilot panel still needs its own filled `outputs/external_predict_results.csv` before Gate 6 can pass |
 | Probability table uncalibrated | All "P(≥1 active) ≈ 92–97%" statements are upper bounds; honest estimate 55–80% for any hit, 35–60% for novel family |
 
 ---

diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md
@@ -266,7 +266,7 @@ review (2026-06-28). Progress on these would materially raise breakthrough proba
 | Gap | Why it matters | Effort estimate |
 |-----|----------------|-----------------|
 | ~~Large-scale benchmark (≥ 500 AMPs vs composition-matched decoys, cluster-split)~~ | ~~Current AUROC 0.8420 measured on 43+44 demo set (n=87, CI₉₅: 0.76–0.91); may not generalise~~ | **Partial** (PR #110: expanded to 95 AMPs + 96 decoys, n=191, AUROC=0.7832, CI₉₅: 0.72–0.84; 52 new public-domain AMPs from 12 taxonomic classes; covers defensins, proline-rich, lantibiotics — a more honest estimate. Cluster-split benchmark added: cluster-aware CI 0.7061–0.8526, representative AUROC 0.7607. 500+ target still deferred to v1.0+) |
-| External predictor ensemble adapters (CAMPR4, AMPScanner, dbAMP, AntiCP2, Macrel) | Independent second opinions on activity; required for scientific credibility. Manual web-submission checklist at `outputs/external_predict_checklist.md`. Macrel v1.6.0 CLI ONNX bug documented PR #77 — all sequences (incl. canonical AMPs magainin-2, LL-37) misclassified as NAMP; use web server at big-data-biology.org/software/macrel. AMPlify omitted: GPU/ONNX env incompatible with current deps | Medium (checklist generated; web submissions pending) |
+| External predictor ensemble adapters (CAMPR4, AMPScanner, dbAMP, AntiCP2, Macrel) | Independent second opinions on activity; required for scientific credibility. Wave 0.5 has completed external evidence for AMPScanner v2, AMPActiPred, Macrel web, HemoFinder, and AntiCP 2.0 with CAMPR4 excluded; see `docs/METRICS_CURRENT.md`. The reusable generic 5-tool pilot-panel workflow still requires a filled `outputs/external_predict_results.csv` per future panel. Macrel v1.6.0 CLI ONNX bug documented PR #77 — all sequences (incl. canonical AMPs magainin-2, LL-37) misclassified as NAMP; use web server at big-data-biology.org/software/macrel. AMPlify omitted: GPU/ONNX env incompatible with current deps | Partial (Wave 0.5 complete; generic future-panel Gate 6 remains panel-specific) |
 | ~~True novelty check against APD3, DRAMP v3.0, dbAMP~~ | ~~Current novelty scored against 45-sequence seed set only; may overestimate novelty~~ | **Done** (v0.5.7: BioPython BLOSUM62 local alignment vs 27,234 AMPs from APD6+DRAMP+UniProt; Wave 0.5 novelty corrected from 53/60 RELATED_NOVEL → 39 CLOSE_RELATIVE + 19 KNOWN_VARIANT + 1 RELATED_NOVEL) |
 | ~~AUPRC alongside AUROC~~ | ~~Better metric for class-imbalanced AMP datasets~~ | **Done** (PR #58; updated PR #72) — pipeline AUPRC = 0.8627 |
 | Wet-lab result integration (active-learning round 2) | Required to move from 15–30% to 50%+ credible probability | Requires wet-lab |

diff --git a/tests/test_docs_consistent.py b/tests/test_docs_consistent.py
@@ -181,3 +181,27 @@ def test_doc_wave05_historical_docs_do_not_pose_as_live_pending_state(self):
         assert "Historical placeholder; external predictor review later completed" in prereg_text
         assert "The external predictor portion has since been completed" in checklist_text
         assert "External predictor review for Wave 0.5 was completed after this panel recommendation was first drafted." in wave1_text
+
+    def test_doc_external_predictor_gate_distinguishes_generic_from_wave05(self):
+        """Live docs must not imply the completed Wave 0.5 screen is still wholly pending."""
+        consensus_text = (DOCS_DIR / "EXTERNAL_PREDICTOR_CONSENSUS.md").read_text(
+            encoding="utf-8"
+        )
+        decision_text = (DOCS_DIR / "DECISION_RULES.md").read_text(encoding="utf-8")
+        reviewer_text = (DOCS_DIR / "REVIEWER_SUMMARY.md").read_text(encoding="utf-8")
+        roadmap_text = (DOCS_DIR / "ROADMAP.md").read_text(encoding="utf-8")
+        expert_text = (DOCS_DIR / "EXPERT_REVIEW_PACK.md").read_text(encoding="utf-8")
+
+        assert "generic pilot-panel consensus workflow" in consensus_text
+        assert "Wave 0.5 activity consensus" in consensus_text
+        assert "three activity predictors" in consensus_text
+        assert "generic 5-tool" in consensus_text
+        assert "Wave 0.5 remains wholly pending" in consensus_text
+        assert "PENDING (see `outputs/external_predict_checklist.md`)" not in decision_text
+        assert "generic Gate 6" in decision_text
+        assert "completed Wave 0.5 external" in decision_text
+        assert "No external predictor ensemble yet" not in reviewer_text
+        assert "Wave 0.5 external screen is complete" in reviewer_text
+        assert "web submissions pending" not in roadmap_text
+        assert "Wave 0.5 complete; generic future-panel Gate 6 remains panel-specific" in roadmap_text
+        assert "CAMPR4 was excluded" in expert_text