docs,test(dogfood): record repair evidence and stabilize installer CI

examples/phase2-final/dogfood-round2/README.md

@@ -0,0 +1,66 @@
+# Dogfood Round 2 repair evidence
+
+This directory records the local Round 2 repair of the failed Phase 2
+dogfood trial. It contains exported, non-canonical evidence only. Canonical
+Benchwork state was changed through the typed MCP boundary and no
+`.benchwork/` state file is included here.
+
+## Root cause
+
+Round 1 attempted to complete `EX-001` with a run plan and metric that did not
+match frozen Protocol `PT-001`. The Protocol required `RUN-000` and metric
+`score`, while the trial recorded `RUN-001` through `RUN-005` using
+`replay_success`. Athanor rejected pilot completion with
+`VALIDATION_REJECTED`, and the failed trial was preserved.
+
+## Repair
+
+The repair preserved all Round 1 records. In particular, frozen `PT-001`,
+active `WK-001`, `EX-001`, and their Chronicle histories remain unchanged.
+The shared `RP-001`, `CL-002`, and `HY-002` aggregate objects advanced through
+the registered evaluation while the repair added a new path:
+
+```text
+PT-002 -> WK-002 -> AR-002 -> EX-002
+       -> RUN-101..RUN-105 -> RB-001 -> AS-001 -> DE-001
+```
+
+- `PT-002` freezes the included Pilot set as `RUN-101` through `RUN-104`.
+- `RUN-105` remains a failed, excluded negative result with an explicit reason.
+- `RB-001` analyzes only the four completed, included runs.
+- `AS-001` records the synthetic-host limitations.
+- `DE-001` seals the user-confirmed `REPAIR` outcome.
+- `WK-002` finishes as `COMPLETED`; the original `WK-001` and `EX-001`
+  remain in their failed-trial states.
+
+The paired `replay_success` comparison produced a treatment mean of `1`, a
+control mean of `0`, and a mean difference of `1` with a 95% Student-t
+interval of `[1, 1]`. The standardized effect is unavailable because paired
+differences have zero variance.
+
+## Integrity checks
+
+The final Deep Doctor report passes with 53 verified Chronicle events and 37
+replayable objects. A fresh process restart exported the same status bytes
+before and after restart:
+
+```text
+2d586f3dbb364808da96c1b212bb3371735a022cbdb0c9a226434f6c08aa64a2
+```
+
+Run the local evidence checks from the repository root:
+
+```bash
+sha256sum --check examples/phase2-final/dogfood-round2/SHA256SUMS
+cmp \
+  examples/phase2-final/dogfood-round2/round2-before-restart.json \
+  examples/phase2-final/dogfood-round2/round2-after-restart.json
+jq -e \
+  '.ok and .chronicle_verified and .all_objects_replayable' \
+  examples/phase2-final/dogfood-round2/restart-doctor.json
+```
+
+The Round 1 Protocol and Experiment traces are included so preservation can be
+audited independently of the final aggregate snapshots. Full interactive
+command logs are intentionally omitted because they are noisy and are not
+required to verify the registered result.

examples/phase2-final/dogfood-round2/SHA256SUMS

@@ -0,0 +1,9 @@
+ed3e2d99a7ae7e32f19d1bbacebd73e88c6c2fa8f2c50cd3e2091158446d92ec  examples/phase2-final/dogfood-round2/analysis-spec-round2.json
+3e355f4d64d22dd25adfa215463b0a6d26b499249229496a26a1d34759cfefe5  examples/phase2-final/dogfood-round2/implementation-artifact-round2.json
+ebc306d07eee42cc433829814bef1b6d5c69f2def0a674e170ebde62d926f331  examples/phase2-final/dogfood-round2/restart-doctor.json
+be29cf6487425fe5a5712c3fb347f061f8f06d32a358359a452d1c1ab94ac832  examples/phase2-final/dogfood-round2/restart-verify.txt
+3579161fda803e46a0b871d0162826fca025e9133d5bc772e913826023613867  examples/phase2-final/dogfood-round2/round1-ex001-trace.json
+8b2e437d43a182e346e6bb745f32eafc6de0472eb71d75687058f67a3e41f79c  examples/phase2-final/dogfood-round2/round1-final-status.json
+b929e98a6730a15bf7a6e13972a2b7224ff198a399c9268f69458d348f9141fe  examples/phase2-final/dogfood-round2/round1-pt001-trace.json
+2d586f3dbb364808da96c1b212bb3371735a022cbdb0c9a226434f6c08aa64a2  examples/phase2-final/dogfood-round2/round2-after-restart.json
+2d586f3dbb364808da96c1b212bb3371735a022cbdb0c9a226434f6c08aa64a2  examples/phase2-final/dogfood-round2/round2-before-restart.json

examples/phase2-final/dogfood-round2/analysis-spec-round2.json

@@ -0,0 +1,35 @@
+{
+  "schema_version": "analysis-spec/1.0",
+  "comparisons": [
+    {
+      "comparison_id": "CMP-002",
+      "experiment_id": "EX-002",
+      "arms": [
+        "baseline",
+        "treatment"
+      ],
+      "metric": "replay_success",
+      "estimand": "mean_difference",
+      "pairing": "paired",
+      "uncertainty_method": "student_t",
+      "confidence_level": 0.95
+    }
+  ],
+  "multiple_comparison_policy": "none",
+  "practical_significance_thresholds": {
+    "replay_success": 0.0
+  },
+  "pilot_run_ids": [
+    "RUN-101",
+    "RUN-102",
+    "RUN-103",
+    "RUN-104"
+  ],
+  "expected_run_ids": [
+    "RUN-101",
+    "RUN-102",
+    "RUN-103",
+    "RUN-104",
+    "RUN-105"
+  ]
+}

examples/phase2-final/dogfood-round2/implementation-artifact-round2.json

@@ -0,0 +1,19 @@
+{
+  "schema_version": "dogfood-implementation-artifact/1.0",
+  "name": "benchwork-dogfood-replay-repair",
+  "version": "0.2",
+  "purpose": "Validate canonical replay through a Protocol-aligned registered pilot and preserve the excluded failed Run.",
+  "protocol_id": "PT-002",
+  "working_id": "WK-002",
+  "analysis_spec": "analysis-spec-round2.json",
+  "experiment_id": "EX-002",
+  "required_pilot_run_ids": [
+    "RUN-101",
+    "RUN-102",
+    "RUN-103",
+    "RUN-104"
+  ],
+  "preserved_failed_run_ids": [
+    "RUN-105"
+  ]
+}

examples/phase2-final/dogfood-round2/restart-doctor.json

@@ -0,0 +1,57 @@
+{
+  "schema_version": "doctor-report/1.1",
+  "mode": "deep",
+  "ok": true,
+  "chronicle_verified": true,
+  "all_objects_replayable": true,
+  "checks": {
+    "chronicle": {
+      "status": "PASS",
+      "event_count": 53
+    },
+    "projection": {
+      "status": "PASS",
+      "object_count": 37
+    },
+    "snapshots": {
+      "status": "PASS",
+      "verified_count": 6
+    },
+    "capsules": {
+      "status": "PASS",
+      "verified_count": 6
+    },
+    "capability_registry": {
+      "status": "PASS",
+      "verified_count": 21
+    },
+    "rite_registry": {
+      "status": "PASS",
+      "verified_count": 3
+    },
+    "grimoire_registry": {
+      "status": "PASS",
+      "verified_count": 0
+    },
+    "agent_outputs": {
+      "status": "PASS",
+      "verified_count": 3
+    },
+    "artifacts": {
+      "status": "PASS",
+      "verified_count": 2
+    },
+    "result_exports": {
+      "status": "PASS",
+      "verified_count": 1
+    },
+    "project_context": {
+      "status": "PASS",
+      "active_program_id": "RP-001"
+    },
+    "migration_state": {
+      "status": "PASS",
+      "residual_count": 0
+    }
+  }
+}

examples/phase2-final/dogfood-round2/restart-verify.txt

@@ -0,0 +1 @@
+Chronicle healthy: 53 verified event(s), receipt chain intact