ops: register Evidence Bench external #9 and 33rd maintenance target

HARNESS_STATUS.md

@@ -5,7 +5,7 @@
 
 A harness may have several independent interventions and PRs. A historical `AWAITING_EXTERNAL` batch state **remains historical** even when a separate publication ledger later confirms an upstream merge. All external submission/acceptance claims below come from the linked owning source; `SUBMITTED_UNVERIFIED_LIVE` means **no current PR-state claim**.
 
-**Scope:** 153 harnesses · 168 interventions · 39 unique external PR URLs · 25 assessment batches · 32 monitored PR targets.
+**Scope:** 153 harnesses · 169 interventions · 40 unique external PR URLs · 25 assessment batches · 33 monitored PR targets.
 
 ## Harness overview
 
@@ -81,7 +81,7 @@ A harness may have several independent interventions and PRs. A historical `AWAI
 | [k4yt3x/meka](https://github.com/k4yt3x/meka) | 1 | 0 | 0 | 0 | 0 |
 | [khanuzeb/axiom](https://github.com/khanuzeb/axiom) | 1 | 0 | 0 | 0 | 0 |
 | [knowflow-ai/KnowWork](https://github.com/knowflow-ai/KnowWork) | 2 | 1 | 0 | 0 | 0 |
-| [kstawiski/evidence-gated-scientific-agent](https://github.com/kstawiski/evidence-gated-scientific-agent) | 1 | 1 | 0 | 0 | 0 |
+| [kstawiski/evidence-gated-scientific-agent](https://github.com/kstawiski/evidence-gated-scientific-agent) | 2 | 1 | 0 | 0 | 0 |
 | [KumihoIO/Revka](https://github.com/KumihoIO/Revka) | 1 | 1 | 0 | 0 | 1 |
 | [kushagrathisside/bridle](https://github.com/kushagrathisside/bridle) | 1 | 0 | 0 | 0 | 0 |
 | [lbdremy/phax](https://github.com/lbdremy/phax) | 2 | 1 | 0 | 0 | 0 |
@@ -245,6 +245,7 @@ A harness may have several independent interventions and PRs. A historical `AWAI
 | knowflow-ai/KnowWork | [assessment:knowwork-sandbox-multiprocess-ownership](data/assessment-candidate-batches/2026-11-10-25.json) | S2 | AWAITING_EXTERNAL | NOT_SUBMITTED | — | assessment-upstream-2026-11-10-25 | Refresh source; check external publication grant |
 | knowflow-ai/KnowWork | [monitor:knowflow-ai/knowwork#4](data/pr-maintenance/targets.json) | — | NOT_TRACKED | WATCHLIST_ONLY | [#4](https://github.com/knowflow-ai/KnowWork/pull/4) | monitor_only | Read PR maintenance inbox; no technical inference |
 | kstawiski/evidence-gated-scientific-agent | [assessment:evidence-bench-native-evidence-admission](data/assessment-candidate-batches/2026-10-23-11.json) | S1 | AWAITING_EXTERNAL | NOT_SUBMITTED | — | assessment-upstream-2026-10-23-11 | Refresh source; check external publication grant |
+| kstawiski/evidence-gated-scientific-agent | [monitor:kstawiski/evidence-gated-scientific-agent#9](data/pr-maintenance/targets.json) | — | NOT_TRACKED | WATCHLIST_ONLY | [#9](https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9) | monitor_only | Read PR maintenance inbox; no technical inference |
 | KumihoIO/Revka | [assessment:revka-workflow-validation-contract](data/assessment-candidate-batches/2026-10-23-10.json) | S1 | AWAITING_EXTERNAL | BLOCKED_PUBLICATION | — | assessment-upstream-2026-10-23-10 | Resolve documented publication gate |
 | kushagrathisside/bridle | [assessment:bridle-run-admission-crash-recovery](data/assessment-candidate-batches/2026-11-09-19.json) | S2 | CLOSED_NO_CHANGE | NONE | — | assessment-upstream-2026-11-09-19 | No code change justified |
 | lbdremy/phax | [assessment:phax-cross-run-overlap-reconciliation](data/assessment-candidate-batches/2026-11-06-13.json) | S2 | AWAITING_EXTERNAL | NOT_SUBMITTED | — | assessment-upstream-2026-11-06-13 | Refresh source; check external publication grant |

data/external-pr-publication-readbacks/2026-11-12.json

@@ -0,0 +1,23 @@
+{
+  "schema_version": 1,
+  "observed_at": "2026-11-12",
+  "source": "Authenticated dev-alpha-2 GitHub REST POST and exact external/fork-local GET readback; effective standing publication mandate",
+  "authenticated_actor": "xLagerFeuer",
+  "metric_boundary": "Publication evidence only; no maintainer merge, verified adoption or VSM regrade",
+  "cases": [
+    {
+      "id": "evidence-bench-native-evidence-admission",
+      "repository": "kstawiski/evidence-gated-scientific-agent",
+      "public_fork": "opensiro-contributions/evidence-gated-scientific-agent",
+      "fork_local_pr_url": "https://github.com/opensiro-contributions/evidence-gated-scientific-agent/pull/1",
+      "pr_url": "https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9",
+      "published_at": "2026-11-12T00:11:59Z",
+      "base_sha": "85b396a824dfad6ec0880cbe07d86d3e24920ae8",
+      "head_sha": "9710d908adb05bc5effc6a557de321bff8cda142",
+      "pr_commits": 1,
+      "publication_state": "OPEN_READY",
+      "approved_exact_title_body": true,
+      "approval_boundary": "ACTIVE standing external PR mandate; authenticated xLagerFeuer dev-alpha-2 REST POST and exact title/body/head/base/two-file external readback; 16/16 Python policy tests PASS and full Python non-live test command exited 0; npm ci not run (Node/npm unavailable), disclosed in external PR. No legal-person CLA/DCO or maintainer acceptance claim."
+    }
+  ]
+}

data/pr-maintenance/targets.json

@@ -111,6 +111,11 @@
       "number": 4,
       "url": "https://github.com/knowflow-ai/KnowWork/pull/4"
     },
+    {
+      "repository": "kstawiski/evidence-gated-scientific-agent",
+      "number": 9,
+      "url": "https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9"
+    },
     {
       "repository": "lbdremy/phax",
       "number": 205,

docs/PR_MAINTENANCE.md

@@ -34,7 +34,7 @@ This is a construction/operating contract. The read-only detector and alert tran
 
 The watcher uses Python standard library only and authenticates using the GitHub Actions token. Each observation is a live **GET** from the GitHub repository that owns it; unexpected API errors, missing identities, unknown states and truncated pagination **fail closed**. Third-party message **bodies are deliberately not stored, executed, or trusted as instructions**: only event IDs, author, status and GitHub URLs are used. Humans/agents inspect the exact review on the external source before repairing.
 
-Current pinned targets are **32** reviewed external PR identities for contributor `xLagerFeuer` (18 previously retained, seven 2026-11-10 publication readbacks, [KnowWork #4](https://github.com/knowflow-ai/KnowWork/pull/4), [Runloop #1](https://github.com/murilo-bassan/durable-agent-runtime/pull/1), [CodeAct #26](https://github.com/xingyaoww/code-act/pull/26), [Clouds-Coder #50](https://github.com/FonaTech/Clouds-Coder/pull/50), [yoagent #450](https://github.com/yologdev/yoagent/pull/450), and [phax #205](https://github.com/lbdremy/phax/pull/205) and [Fara #96](https://github.com/microsoft/fara/pull/96) on 2026-11-11), including Headcount #47 (previously verified but returning GitHub 404 on 2026-11-05), closed-unmerged StarNet #56, and six earlier Corrective PRs: [Workflow One #173](https://github.com/chumingjun/dsh-harness-one/pull/173) (merged 2026-10-29), [Reef/Polyp #19](https://github.com/Coral-Bricks-AI/reef/pull/19), [Spartan #1](https://github.com/CorticalComputer/Spartan/pull/1), [BESSER #205](https://github.com/BESSER-PEARL/BESSER-Agentic-Framework/pull/205), [BabyAI #128](https://github.com/mila-iqia/babyai/pull/128), and [Phoenix #1](https://github.com/Phoenixrr2113/agent-harness/pull/1). This list is a versioned scope snapshot and is **not** dynamically widened by an unreviewed search result. Adding/removing targets requires a reviewable manifest change and explicit checkpoint-migration handling. Closed/merged PRs remain in the snapshot to preserve lifecycle evidence until a deliberate archival change.
+Current pinned targets are **33** reviewed external PR identities for contributor `xLagerFeuer` (18 previously retained, seven 2026-11-10 publication readbacks, [KnowWork #4](https://github.com/knowflow-ai/KnowWork/pull/4), [Runloop #1](https://github.com/murilo-bassan/durable-agent-runtime/pull/1), [CodeAct #26](https://github.com/xingyaoww/code-act/pull/26), [Clouds-Coder #50](https://github.com/FonaTech/Clouds-Coder/pull/50), [yoagent #450](https://github.com/yologdev/yoagent/pull/450), and [phax #205](https://github.com/lbdremy/phax/pull/205) and [Fara #96](https://github.com/microsoft/fara/pull/96) on 2026-11-11), including Headcount #47 (previously verified but returning GitHub 404 on 2026-11-05), closed-unmerged StarNet #56, and six earlier Corrective PRs: [Workflow One #173](https://github.com/chumingjun/dsh-harness-one/pull/173) (merged 2026-10-29), [Reef/Polyp #19](https://github.com/Coral-Bricks-AI/reef/pull/19), [Spartan #1](https://github.com/CorticalComputer/Spartan/pull/1), [BESSER #205](https://github.com/BESSER-PEARL/BESSER-Agentic-Framework/pull/205), [BabyAI #128](https://github.com/mila-iqia/babyai/pull/128), and [Phoenix #1](https://github.com/Phoenixrr2113/agent-harness/pull/1). This list is a versioned scope snapshot and is **not** dynamically widened by an unreviewed search result. Adding/removing targets requires a reviewable manifest change and explicit checkpoint-migration handling. Closed/merged PRs remain in the snapshot to preserve lifecycle evidence until a deliberate archival change.
 
 **Reviewed availability-only exception (Upstream #139 / PR #140):** Headcount #47 is retained as the eighteenth identity, with the sole allowlisted `monitoring: verify_404_only` manifest mode. Every run MUST GET that exact source and confirm HTTP 404. It may then report **17 fully observed / 1 explicitly UNVERIFIABLE of 18**, continue reading the other 17, and store a distinct *availability marker* instead of stale PR fields. The first transition emits one inbox `SOURCE_UNVERIFIABLE` alert, never a fabricated close/merge/review. Recovery (HTTP 200), a different status, any additional unavailable URL, or a malformed source **fails closed** pending source review and removal of the exception. The state is never silently `OPEN` or `CLOSED_UNMERGED`. This is a reviewed change to monitoring *coverage*, not a removal of the historical cohort, or permission to claim an 18-of-18 event snapshot. A returned count of 17 is prominently labelled in the observer and inbox output.
 
@@ -83,3 +83,16 @@ were not executed. This exact allowlisted addition permits a 31 to 32
 maintenance-target migration without replaying historical comments or
 inventing maintainer activity. Publication is not third-party acceptance,
 measured downstream benefit or a VSM regrade.
+
+**Checkpoint migration for Evidence Bench #9 (2026-11-12 UTC, standing publication mandate):**
+External [Evidence Bench #9](https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9)
+was published Open/Ready from [fork-local #1](https://github.com/opensiro-contributions/evidence-gated-scientific-agent/pull/1),
+on original main at 85b396a824dfad6ec0880cbe07d86d3e24920ae8, source-native
+branch head 9710d908adb05bc5effc6a557de321bff8cda142; two changed files.
+Native Python policy regression 16/16 PASS, full Python non-live suite command
+exited 0 after uv sync frozen/dev/analysis; no test total separately recorded.
+The contributor's documented npm ci step was not performed, with Node/npm
+absent, disclosed in external body. This single allowlisted addition enables
+32 to 33 maintenance targets, preserving earlier checkpoint state and recording
+only NEW_TRACKED_PR informational transitions. No maintainer merge, real-world
+benefit, or VSM-grade change is inferred.

scripts/pr_maintenance_watch.py

@@ -64,6 +64,8 @@
     "https://github.com/FonaTech/Clouds-Coder/pull/50",
     # yoagent standing-mandate publication, native all-features validated on 2026-11-11.
     "https://github.com/yologdev/yoagent/pull/450",
+    # Evidence Bench policy 16/16 and full non-live exit 0; external #9 readback 2026-11-12.
+    "https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9",
     # Fara focused 3/3 native tests and exact external #96 readback 2026-11-11.
     "https://github.com/microsoft/fara/pull/96",
     # phax source-native full-gate PASS and standing publication 2026-11-11.

tests/test_portfolio_reconcile.py

@@ -50,9 +50,9 @@ def pages(self, path):
 class PortfolioReconcileTests(unittest.TestCase):
     def test_frozen_source_union_exact_size_and_identity(self):
         entries = portfolio.cohort()
-        self.assertEqual(len(entries), 39)
-        self.assertEqual(sum("watchlist" in row["groups"] for row in entries), 32)
-        self.assertEqual(len({row["url"] for row in entries}), 39)
+        self.assertEqual(len(entries), 40)
+        self.assertEqual(sum("watchlist" in row["groups"] for row in entries), 33)
+        self.assertEqual(len({row["url"] for row in entries}), 40)
         self.assertEqual(len([row for row in entries if "casebook" in row["groups"]]), 9)
         self.assertEqual(len([row for row in entries if "registry" in row["groups"]]), 1)
         self.assertEqual(

tests/test_pr_maintenance_watch.py

@@ -145,10 +145,10 @@ def write_targets(self, root):
                                 "targets": [{"repository": "Example/project", "number": 3, "url": URL}]}))
         return p
 
-    def test_exact_manifest_has_thirty_two_targets_and_no_duplicates(self):
+    def test_exact_manifest_has_thirty_three_targets_and_no_duplicates(self):
         data = watch.validate_targets()
-        self.assertEqual(len(data["targets"]), 32)
-        self.assertEqual(len({t["url"] for t in data["targets"]}), 32)
+        self.assertEqual(len(data["targets"]), 33)
+        self.assertEqual(len({t["url"] for t in data["targets"]}), 33)
         self.assertIn("https://github.com/knowflow-ai/KnowWork/pull/4",
                       {t["url"] for t in data["targets"]})
         self.assertIn("https://github.com/knowflow-ai/KnowWork/pull/4",
@@ -165,6 +165,10 @@ def test_exact_manifest_has_thirty_two_targets_and_no_duplicates(self):
         self.assertIn("https://github.com/yologdev/yoagent/pull/450", watch.APPROVED_CHECKPOINT_ADDITIONS)
         self.assertIn("https://github.com/FonaTech/Clouds-Coder/pull/50", {t["url"] for t in data["targets"]})
         self.assertIn("https://github.com/FonaTech/Clouds-Coder/pull/50", watch.APPROVED_CHECKPOINT_ADDITIONS)
+        self.assertIn("https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9",
+                      {t["url"] for t in data["targets"]})
+        self.assertIn("https://github.com/kstawiski/evidence-gated-scientific-agent/pull/9",
+                      watch.APPROVED_CHECKPOINT_ADDITIONS)
         self.assertIn("https://github.com/microsoft/fara/pull/96",
                       {t["url"] for t in data["targets"]})
         self.assertIn("https://github.com/microsoft/fara/pull/96",

tests/test_resolved_pr_acceptance.py

@@ -139,7 +139,7 @@ def test_multiple_bot_report_comments_fail_before_write(self):
         self.assertFalse(api.writes)
 
     def test_only_reviewed_watchlist_and_no_unreviewed_expansion(self):
-        self.assertEqual(len(metric.validate_targets()["targets"]), 32)
+        self.assertEqual(len(metric.validate_targets()["targets"]), 33)
         self.assertEqual(metric.INBOX_REPO, "opensiro/upstream")
         self.assertEqual(metric.INBOX_NUMBER, 125)
 

tests/test_verified_external_outcomes.py

@@ -87,12 +87,12 @@ class VerifiedExternalOutcomeTests(unittest.TestCase):
     def test_owner_ledgers_cover_exact_portfolio_with_one_replacement_chain(self):
         items, blocked = outcomes.cohort()
         self.assertEqual((len(items), sum(len(x["pr_urls"]) for x in items), blocked),
-                         (38, 39, 21))
+                         (39, 40, 21))
         self.assertEqual(len([x for x in items if x["kind"] == "improvement"]), 2)
         self.assertEqual(len([x for x in items if x["kind"] == "corrective"]), 13)
         self.assertEqual(len([x for x in items if x["kind"] == "casebook"]), 8)
         self.assertEqual(len([x for x in items if x["kind"] == "registry"]), 1)
-        self.assertEqual(len([x for x in items if x["kind"] == "publication"]), 14)
+        self.assertEqual(len([x for x in items if x["kind"] == "publication"]), 15)
         self.assertIn("publication:knowwork-sandbox-start-race",
                       {x["id"] for x in items})
         self.assertIn("publication:runloop-budget-finalization-idempotency",