experiment(exp64): merged rule ownership real-Aegis red → green

.github/workflows/exp64-evidence.yml

@@ -0,0 +1,46 @@
+name: exp64-evidence
+on:
+  pull_request:
+  push:
+    branches: [main]
+jobs:
+  replay-and-real-package:
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    steps:
+      - uses: actions/checkout@v4
+      - uses: pnpm/action-setup@v4
+        with:
+          version: 10.9.0
+      - uses: actions/setup-node@v4
+        with:
+          node-version: '22'
+          cache: npm
+          cache-dependency-path: experiments/64-merged-rule-ownership/ci/package-lock.json
+      - name: Install portable core/experiment toolchain
+        # Older experiments retain machine-local file deps; do not rewrite unrelated packages.
+        run: |
+          npm ci --prefix experiments/64-merged-rule-ownership/ci --workspaces=false
+          ln -s experiments/64-merged-rule-ownership/ci/node_modules node_modules
+          mkdir -p node_modules/@swarmlab
+          ln -s "$GITHUB_WORKSPACE/core" node_modules/@swarmlab/core
+      - name: Build and typecheck core plus new experiment
+        run: |
+          node_modules/.bin/tsc -b core experiments/64-merged-rule-ownership
+          node_modules/.bin/tsc -b --force core experiments/64-merged-rule-ownership
+      - name: Verify frozen corpus and exact content-addressed experiment receipts
+        run: npm run verify:evidence && node experiments/64-merged-rule-ownership/verify-receipts.mjs
+      - name: Checkout exact candidate (not moving main)
+        uses: actions/checkout@v4
+        with:
+          repository: heybeaux/aegis
+          ref: edb9f557fc776b5710360a453f0df0146f3b977b
+          path: .aegis-candidate
+      - name: Build real Aegis
+        working-directory: .aegis-candidate
+        run: pnpm install --frozen-lockfile && pnpm --filter @heybeaux/lattice-aegis build
+      - name: Execute same frozen experiment and inspect rows/metrics/actual source-dist receipts
+        run: |
+          AEGIS_REPO="$GITHUB_WORKSPACE/.aegis-candidate" node experiments/64-merged-rule-ownership/dist/main.js > "$RUNNER_TEMP/exp64.log"
+          cat "$RUNNER_TEMP/exp64.log"
+          node experiments/64-merged-rule-ownership/verify-receipts.mjs --run "$RUNNER_TEMP/exp64.log"

CLAIMS.json

@@ -6333,5 +6333,75 @@
     "evidenceStatus": "verified",
     "seedSetId": "async-evaluator-config-integrity-v1",
     "notes": "Pre-registration b0378e0 BEFORE implementation/output; 21 frozen scenarios. Current-main baseline repeated with identical 63 arm scenario rows, same harness hash candidate; full 68-event replay equality. Two diagnostic traces had fixture-only I/O-accounting error; preserved locally but excluded from claims. Thresholds/scenarios unchanged. Exp57/RT47 remains prior unlanded work, NOT novelty tonight. No threshold tuned; holdout reserved unused. Authentic plain-data host, no post-return/ABA/getter/proxy/latency claims."
+  },
+  {
+    "id": "RT-54",
+    "claim": "Mutable caller recipes must not silently edit a constructed effective cascade or receive consumer edits through merged rule aliases.",
+    "experiment": "experiments/64-merged-rule-ownership",
+    "spec": "specs/70-merged-rule-ownership.md",
+    "runIds": [
+      "mro-mv96376j",
+      "mro-mv9637ag",
+      "mro-mv966i1f"
+    ],
+    "tracePaths": [
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl"
+    ],
+    "scoreFields": [
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.detachedControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisProducerDriftRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisUnsafeAllowRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisResolutionAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisConsumerIsolation",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisCallerOwnership",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.detachedControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisProducerDriftRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisUnsafeAllowRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisResolutionAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisConsumerIsolation",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisCallerOwnership",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.detachedControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisProducerDriftRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisUnsafeAllowRate",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisResolutionAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisConsumerIsolation",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisControlAccuracy",
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisCallerOwnership"
+    ],
+    "expectedValues": {
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.detachedControlAccuracy": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisProducerDriftRate": 0.8888888888888888,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisUnsafeAllowRate": 0.8888888888888888,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisResolutionAccuracy": 0.38095238095238093,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisConsumerIsolation": 0,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisControlAccuracy": 0.7777777777777778,
+      "experiments/64-merged-rule-ownership/runs/mro-mv96376j.jsonl#last.aegisCallerOwnership": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.detachedControlAccuracy": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisProducerDriftRate": 0.8888888888888888,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisUnsafeAllowRate": 0.8888888888888888,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisResolutionAccuracy": 0.38095238095238093,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisConsumerIsolation": 0,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisControlAccuracy": 0.7777777777777778,
+      "experiments/64-merged-rule-ownership/runs/mro-mv9637ag.jsonl#last.aegisCallerOwnership": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.detachedControlAccuracy": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisProducerDriftRate": 0,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisUnsafeAllowRate": 0,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisResolutionAccuracy": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisConsumerIsolation": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisControlAccuracy": 1,
+      "experiments/64-merged-rule-ownership/runs/mro-mv966i1f.jsonl#last.aegisCallerOwnership": 1
+    },
+    "reproductionCommand": "npm run build && AEGIS_REPO=/path/to/pinned/real/aegis node experiments/64-merged-rule-ownership/dist/main.js",
+    "stackRecommendation": "Each returned effective Rule owns a structured clone; precedence, warnings and mutable output contract unchanged.",
+    "stackReposAndPRs": [
+      "https://github.com/heybeaux/aegis/pull/70",
+      "https://github.com/heybeaux/swarmlab/pull/53"
+    ],
+    "evidenceStatus": "verified",
+    "notes": "Unique synchronous cascade value-ownership boundary. Baseline origin/main4be84ed repeated identical rows, 8/9 unsafe allow; f274278 unchanged-harness21/21 green. Holdout unused (no tuning), raw traces/package manifests preserved. Candidate not remotely merged; RT50\u201353 pending and not imported. Direct result mutation/loadPack aliasing unproven by design."
   }
 ]

JOURNAL.md

@@ -307,3 +307,52 @@ Spec64 b0378e0; 21 frozen scenarios, seed async-evaluator-config-integrity-v1. R
 Local regression:1087 Aegis tests, gate47/47; SwarmLab46 claims/96 traces/881 assertions; exp55/56 green regression only. Orphaned exp57 preserved, not landed or independently certified. RT47 rebase must reconcile trusted clock function capture with RT48 plain-policy snapshot; combined behavior unproven.
 
 Linked PRs: https://github.com/heybeaux/aegis/pull/63 (Aegis), https://github.com/heybeaux/swarmlab/pull/46 (SwarmLab). Local green; CI status inspected in tonight's pipeline ledger before any merge.
+
+## RT54 — Merged rule reference ownership (exp64 / Spec70, 2026-10-14)
+
+New pre-registered question: can caller-owned rule recipes edit already-constructed
+cascade policy between synchronous merge and compilation? Pre-registration `4e5c4c8`;
+21 frozen scenarios, no tuning or LLM judge. Distinct from in-flight RT48, receipt
+metadata RT49, defaults RT51, compiled-cache RT52 and overlay consent RT53.
+
+Real Aegis `4be84ed` baseline `mro-mv96376j` / `mro-mv9637ag` identical rows:
+producer drift **8/9**, unsafe allow **8/9**, exact resolution **8/21**;
+consumer isolation **0**, compatibility controls **7/9**, caller ownership **1**.
+Independent detached-boundary diagnostic **21/21** on both baselines; it is a
+control, not replacement Aegis. These red traces remain committed.
+
+Runtime patch `f27427879984856bad36731788376dbc74702aaa` clones each effective Rule independently at
+merge return. Separate cloning is important: cloning the whole array preserves
+shared nested identities across distinct rule ids. Same unchanged harness/seeds,
+`mro-mv966i1f`: producer drift **0/9**, unsafe allow **0/9**, resolution **21/21**,
+all secondary metrics **1**,46-event full replay; manifests pin real package.
+Focused frozen tests **16 red/4 pass ->20 green**. Thresholds unchanged.
+
+Host owns provenance, deliberate policy edits and reloading; result remains
+editable. Direct loadPack binding, live immutability/atomic execution, getters or
+proxies are not claimed. Stored scores are historical regression evidence,
+not runtime verification. Full regression and remote CI/PR status will be recorded
+below when actually observed. Prior exp60–63 candidates remain separate/unmerged.
+
+### Regression verification (not new evidence)
+Full Aegis release:check at `edb9f55`: build/typecheck **1145/1145 tests**
+(628core,152Aegis,22collect,86label,203hook,54bench), static measured evidence **50/50**.
+Initial evidence catalogue update exposed one stale expected report banner, not a
+runtime regression: corrected RT49->RT54 expectation; preserved red log, full rerun
+green. SwarmLab build/typecheck and verify:evidence **49claims105traces980assertions**.
+Fresh same-harness `mro-mv96dcb4` on clean final candidate **21/21**, actual source/dist
+manifest audited. Frozen red+green46-event receipts verified; no thresholds moved.
+Baseline frozen/remerge compatibility controls fail on real old Aegis7/9, whereas
+the detached diagnostic is21/21, distinguishing general ownership gap from fixture
+failure. Holdout remains unused. No source .acr changes; Territory not refreshed.
+
+### Linked PRs and observed GitHub CI
+Aegis [PR70](https://github.com/heybeaux/aegis/pull/70), candidate
+`edb9f557fc776b5710360a453f0df0146f3b977b`: release-check
+[38694629492](https://github.com/heybeaux/aegis/actions/runs/38694629492) **SUCCESS**.
+SwarmLab [PR53](https://github.com/heybeaux/swarmlab/pull/53), evidence
+`b973bd8db231b4168191a773f9f0e06f463192a2`: portable replay + pinned real-package
+[38694637393](https://github.com/heybeaux/swarmlab/actions/runs/38694637393) **SUCCESS**.
+Both inspected CLEAN/MERGEABLE; CodeRabbit success is review-skipped, **not review
+approval**. No protections bypassed. Final documentation commit will get its own CI;
+merge/main SHA/status recorded in the nightly ledger, not inferred here.

SYNTHESIS.md

@@ -1505,3 +1505,40 @@ function overrides and post-return changes unproven. Earlier exp57/RT47 is unlan
 not tonight's novelty; this independent origin/main patch changes no elapsed-clock contract.
 Two initial diagnostic traces had oracle-only I/O counter mismatch (real output unchanged);
 corrected before admitting repeated baseline, preserved but not claimed. Thresholds never moved.
+
+## RT54 — Merged rule reference ownership (exp64 / Spec70, 2026-10-14)
+
+New pre-registered question: can caller-owned rule recipes edit already-constructed
+cascade policy between synchronous merge and compilation? Pre-registration `4e5c4c8`;
+21 frozen scenarios, no tuning or LLM judge. Distinct from in-flight RT48, receipt
+metadata RT49, defaults RT51, compiled-cache RT52 and overlay consent RT53.
+
+Real Aegis `4be84ed` baseline `mro-mv96376j` / `mro-mv9637ag` identical rows:
+producer drift **8/9**, unsafe allow **8/9**, exact resolution **8/21**;
+consumer isolation **0**, compatibility controls **7/9**, caller ownership **1**.
+Independent detached-boundary diagnostic **21/21** on both baselines; it is a
+control, not replacement Aegis. These red traces remain committed.
+
+Runtime patch `f27427879984856bad36731788376dbc74702aaa` clones each effective Rule independently at
+merge return. Separate cloning is important: cloning the whole array preserves
+shared nested identities across distinct rule ids. Same unchanged harness/seeds,
+`mro-mv966i1f`: producer drift **0/9**, unsafe allow **0/9**, resolution **21/21**,
+all secondary metrics **1**,46-event full replay; manifests pin real package.
+Focused frozen tests **16 red/4 pass ->20 green**. Thresholds unchanged.
+
+Host owns provenance, deliberate policy edits and reloading; result remains
+editable. Direct loadPack binding, live immutability/atomic execution, getters or
+proxies are not claimed. Stored scores are historical regression evidence,
+not runtime verification. Full regression and remote CI/PR status will be recorded
+below when actually observed. Prior exp60–63 candidates remain separate/unmerged.
+
+### Linked PRs and observed GitHub CI
+Aegis [PR70](https://github.com/heybeaux/aegis/pull/70), candidate
+`edb9f557fc776b5710360a453f0df0146f3b977b`: release-check
+[38694629492](https://github.com/heybeaux/aegis/actions/runs/38694629492) **SUCCESS**.
+SwarmLab [PR53](https://github.com/heybeaux/swarmlab/pull/53), evidence
+`b973bd8db231b4168191a773f9f0e06f463192a2`: portable replay + pinned real-package
+[38694637393](https://github.com/heybeaux/swarmlab/actions/runs/38694637393) **SUCCESS**.
+Both inspected CLEAN/MERGEABLE; CodeRabbit success is review-skipped, **not review
+approval**. No protections bypassed. Final documentation commit will get its own CI;
+merge/main SHA/status recorded in the nightly ledger, not inferred here.

docs/EVIDENCE-LEDGER.md

@@ -197,3 +197,52 @@ harnesses each replayed against the combined built Aegis package: 65/65 and 68/6
 respectively, with unsafe allow rate 0 and resolution accuracy 1 in both. Aegis release:check
 passed with 49/49 evidence mappings and combined clock/config unit tests. These integration
 checks do not expand the original experiment scope or certify post-return atomicity.
+
+## RT54 — Merged rule reference ownership (exp64 / Spec70, 2026-10-14)
+
+New pre-registered question: can caller-owned rule recipes edit already-constructed
+cascade policy between synchronous merge and compilation? Pre-registration `4e5c4c8`;
+21 frozen scenarios, no tuning or LLM judge. Distinct from in-flight RT48, receipt
+metadata RT49, defaults RT51, compiled-cache RT52 and overlay consent RT53.
+
+Real Aegis `4be84ed` baseline `mro-mv96376j` / `mro-mv9637ag` identical rows:
+producer drift **8/9**, unsafe allow **8/9**, exact resolution **8/21**;
+consumer isolation **0**, compatibility controls **7/9**, caller ownership **1**.
+Independent detached-boundary diagnostic **21/21** on both baselines; it is a
+control, not replacement Aegis. These red traces remain committed.
+
+Runtime patch `f27427879984856bad36731788376dbc74702aaa` clones each effective Rule independently at
+merge return. Separate cloning is important: cloning the whole array preserves
+shared nested identities across distinct rule ids. Same unchanged harness/seeds,
+`mro-mv966i1f`: producer drift **0/9**, unsafe allow **0/9**, resolution **21/21**,
+all secondary metrics **1**,46-event full replay; manifests pin real package.
+Focused frozen tests **16 red/4 pass ->20 green**. Thresholds unchanged.
+
+Host owns provenance, deliberate policy edits and reloading; result remains
+editable. Direct loadPack binding, live immutability/atomic execution, getters or
+proxies are not claimed. Stored scores are historical regression evidence,
+not runtime verification. Full regression and remote CI/PR status will be recorded
+below when actually observed. Prior exp60–63 candidates remain separate/unmerged.
+
+### Regression verification (not new evidence)
+Full Aegis release:check at `edb9f55`: build/typecheck **1145/1145 tests**
+(628core,152Aegis,22collect,86label,203hook,54bench), static measured evidence **50/50**.
+Initial evidence catalogue update exposed one stale expected report banner, not a
+runtime regression: corrected RT49->RT54 expectation; preserved red log, full rerun
+green. SwarmLab build/typecheck and verify:evidence **49claims105traces980assertions**.
+Fresh same-harness `mro-mv96dcb4` on clean final candidate **21/21**, actual source/dist
+manifest audited. Frozen red+green46-event receipts verified; no thresholds moved.
+Baseline frozen/remerge compatibility controls fail on real old Aegis7/9, whereas
+the detached diagnostic is21/21, distinguishing general ownership gap from fixture
+failure. Holdout remains unused. No source .acr changes; Territory not refreshed.
+
+### Linked PRs and observed GitHub CI
+Aegis [PR70](https://github.com/heybeaux/aegis/pull/70), candidate
+`edb9f557fc776b5710360a453f0df0146f3b977b`: release-check
+[38694629492](https://github.com/heybeaux/aegis/actions/runs/38694629492) **SUCCESS**.
+SwarmLab [PR53](https://github.com/heybeaux/swarmlab/pull/53), evidence
+`b973bd8db231b4168191a773f9f0e06f463192a2`: portable replay + pinned real-package
+[38694637393](https://github.com/heybeaux/swarmlab/actions/runs/38694637393) **SUCCESS**.
+Both inspected CLEAN/MERGEABLE; CodeRabbit success is review-skipped, **not review
+approval**. No protections bypassed. Final documentation commit will get its own CI;
+merge/main SHA/status recorded in the nightly ledger, not inferred here.