test: refresh product os reader quality reference

.gitattributes

@@ -1,2 +1,3 @@
 tests/fixtures/fast_rerun_*/** -text
 experiments/080/cases/*/seed_archive/** -text
+src/multi_agent_brief/evaluation_cases/fixtures/cases/*/workspace/** -text

CHANGELOG.md

@@ -9,6 +9,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **Product OS reader-quality reference package**: the packaged
+  `same_evidence_reader_quality_regression` eval case now generates
+  Quality Panel JSON/summary/HTML plus clean delivery/audit bundle archives,
+  and the docs include a public-safe v0.11.3 reference note. This is an
+  inspectable reference-package regression signal only; it does not claim
+  output-quality improvement, semantic proof, delivery approval, or release
+  authorization.
 - **Same-evidence reader-quality regression pack**: packaged public-safe
   evaluation cases now include a synthetic
   `same_evidence_reader_quality_regression` workspace that holds evidence

README.md

@@ -363,6 +363,7 @@ The goal is not to remove human judgment. The goal is to let humans spend more t
 - [Architecture Status](docs/architecture-status.md)
 - [Roadmap](docs/roadmap.md)
 - [Red lines and anti-patterns](docs/red-lines-and-anti-patterns.md)
+- [Product OS reader-quality reference package](docs/reference-runs/v0.11.3-product-os-reader-quality-reference.md)
 - [Synthetic regression pack](docs/reference-runs/v0.11.1-synthetic-regression-pack.md)
 - [Public solar integration run](docs/reference-runs/v0.7.2-public-solar-integration.zh-CN.md)
 - [Organoid-industry failure study](docs/reference-runs/v0.7.4-organoid-failure-study.zh-CN.md)

README.zh-CN.md

@@ -349,6 +349,7 @@ BriefLoop 想把软件工程里的那套“可追踪、可回滚、可审计、
 - [架构状态](docs/architecture-status.zh-CN.md)
 - [路线图](docs/roadmap.zh-CN.md)
 - [红线与反模式](docs/red-lines-and-anti-patterns.md)
+- [Product OS 读者质量 reference package](docs/reference-runs/v0.11.3-product-os-reader-quality-reference.md)
 - [合成回归包](docs/reference-runs/v0.11.1-synthetic-regression-pack.md)
 - [公开运行摘要](docs/reference-runs/v0.7.2-public-solar-integration.zh-CN.md)
 - [失败研究](docs/reference-runs/v0.7.4-organoid-failure-study.zh-CN.md)

docs/README.md

@@ -46,6 +46,7 @@ tree is fully bilingual.
 - [Runtime recipes](runtime-recipes.md)
 - [Support matrix](support-matrix.md)
 - [Red lines and anti-patterns](red-lines-and-anti-patterns.md)
+- [Pre-release v0.11.3 Product OS reader-quality reference package](reference-runs/v0.11.3-product-os-reader-quality-reference.md)
 - [Pre-release v0.11.1 synthetic regression pack](reference-runs/v0.11.1-synthetic-regression-pack.md)
 - [BriefLoop-090 A-controlled auditable-brief pilot](reference-runs/briefloop-090-a-controlled-pilot.md)
 - [Security](security.md)

docs/reference-runs/v0.11.3-product-os-reader-quality-reference.md

@@ -0,0 +1,101 @@
+# Pre-Release v0.11.3 Product OS Reader-Quality Reference Package
+
+This is a public-safe reference note for the pre-release v0.11.3 reference-run
+refresh line. It documents one inspectable BriefLoop-era Product OS package
+using the packaged synthetic `same_evidence_reader_quality_regression` case.
+
+It is not proof of output quality, semantic correctness, source-support
+sufficiency, management readiness, delivery approval, legal approval, or public
+release authorization.
+
+## Scope
+
+- Product entry shape: `industry-weekly` / `market_weekly` workspace semantics.
+- Case type: packaged public-safe synthetic workspace.
+- Reader artifact: `output/delivery/brief.md`.
+- Audit/control artifacts: Claim Ledger, Atomic Claim Graph, Evidence Span
+  Registry, audit report, scoped gate reports, source appendix trace, Quality
+  Panel JSON, Quality Summary, Quality Panel HTML, and clean bundle manifest.
+- Bundle surfaces: `output/delivery_bundle.zip` contains reader-facing delivery
+  artifacts only; `output/audit_bundle.zip` contains audit/control artifacts.
+- Same-evidence boundary: the fixture holds source/evidence inputs fixed while
+  surfacing reader-quality diagnostics.
+
+The case does not call an LLM, fetch live sources, parse binary documents,
+execute specialist agents, or approve delivery.
+
+## What The Package Shows
+
+- Reader-facing delivery can stay free of internal `[src:...]` claim markers.
+- The audit bundle keeps the trace surfaces that the reader bundle omits.
+- The Quality Panel can surface existing Product OS diagnostics together:
+  materiality/focus selection, reader template conformance, support-calibrated
+  wording, trajectory/guidance projections, and closeout guidance.
+- Warning and incomplete surfaces remain visible instead of being narrated away.
+- Clean delivery/audit archives can be generated from the deterministic bundle
+  manifest.
+
+## Reproduce
+
+From a source checkout or installed package:
+
+```bash
+multi-agent-brief eval-cases validate
+multi-agent-brief eval-cases run --case-id same_evidence_reader_quality_regression --json
+```
+
+The packaged case executes this deterministic sequence:
+
+```text
+state.check
+status.show
+quality.summarize
+packs.bundle
+state.check
+status.show
+```
+
+Expected generated artifacts inside the temporary workspace include:
+
+```text
+output/intermediate/quality_panel.json
+output/intermediate/quality_summary.md
+output/intermediate/quality_panel.html
+output/report_bundle_manifest.json
+output/delivery_bundle.zip
+output/audit_bundle.zip
+```
+
+The fixture itself is tracked under:
+
+```text
+src/multi_agent_brief/evaluation_cases/fixtures/cases/same_evidence_reader_quality_regression/
+```
+
+## Known Warnings
+
+The reference case intentionally keeps several non-authoritative diagnostics
+visible:
+
+- source evidence pack materialization is missing;
+- materiality/focus exclusions require human review;
+- reader template conformance has warning-only findings;
+- support-calibrated wording has warning-only findings.
+
+These warnings are part of the reference package. They should not be hidden or
+rephrased as a pass.
+
+## What This Does Not Show
+
+- It does not prove that BriefLoop writes a better brief.
+- It does not prove that a source semantically supports a claim.
+- It does not prove that reader-quality warnings correlate with human
+  preference.
+- It does not prove that a report is management-ready, disclosure-ready, or
+  externally publishable.
+- It does not use private customer, employer, investor-relations, or commercial
+  source material.
+
+Passing the case means the deterministic reference package and bundle surfaces
+remained inspectable. It is a reference-package regression signal only.
+

src/multi_agent_brief/evaluation_cases/contract.py

@@ -30,6 +30,7 @@
     "gates.show",
     "gates.validate",
     "guidance_manifestation.seed_report",
+    "packs.bundle",
     "provenance.build",
     "provenance.show",
     "provenance.validate",

src/multi_agent_brief/evaluation_cases/fixtures/cases/same_evidence_reader_quality_regression/workspace/output/intermediate/finalize_report.json

@@ -1,5 +1,8 @@
 {
   "audit_binding": {
+    "audit_report_sha256": "aac0fcc4b91db3bc6d18970a06e5c9491c1d11d07fd6f76e80f7e1258c8bd701",
+    "audited_brief_sha256": "72826cb1e7ed20aa7b1b94af24af2d0d4cce2d3e57d22f801055d28ee9ef94ce",
+    "claim_ledger_sha256": "0f2407d58506b314d7c899fa77e692c6c3d01838a118b60ba6016dd917b7dc06",
     "findings": [],
     "status": "pass"
   },
@@ -16,6 +19,9 @@
   "delivery_artifacts": [
     "output/delivery/brief.md"
   ],
+  "delivery_artifact_sha256": {
+    "output/delivery/brief.md": "aec52d7b508b852dda9de6586974ef47ddb9d0162925e8c86f37f3bcd5286328"
+  },
   "duplicate_citation_count": 0,
   "quality_panel_closeout": {
     "status": "recommended"

src/multi_agent_brief/evaluation_cases/fixtures/cases/same_evidence_reader_quality_regression/workspace/output/source_appendix.md

@@ -0,0 +1,12 @@
+# Source Appendix
+
+## Sources
+
+### [S1] Synthetic market report excerpt
+
+- Type: local_file
+- Category: market_report
+- Retrieval source type: local_file
+- Underlying evidence type: media_report
+- Note: Public-safe synthetic source used only for deterministic reference-run fixture coverage.
+

src/multi_agent_brief/evaluation_cases/fixtures/cases/same_evidence_reader_quality_regression/workspace/output/source_appendix_trace.md

@@ -0,0 +1,10 @@
+# Source Appendix Trace
+
+This audit trace maps reader-facing source labels back to synthetic fixture
+source records. It is public-safe and deterministic; it does not prove semantic
+source support.
+
+| Reader label | Source ID | Claim ID | Source title |
+|---|---|---|---|
+| S1 | SYN_SRC_001 | SYN_CLAIM_001 | Synthetic market report excerpt |
+

src/multi_agent_brief/evaluation_cases/fixtures/manifest.yaml

@@ -373,6 +373,8 @@ cases:
         args: {}
       - action: quality.summarize
         args: {}
+      - action: packs.bundle
+        args: {}
       - action: state.check
         args: {}
       - action: status.show
@@ -390,6 +392,14 @@ cases:
             quality_panel: output/intermediate/quality_panel.json
             quality_summary: output/intermediate/quality_summary.md
             quality_panel_html: output/intermediate/quality_panel.html
+        - action: packs.bundle
+          exit_code: 0
+          result_contains:
+            report_bundle_manifest: output/report_bundle_manifest.json
+            delivery_bundle_archive: output/delivery_bundle.zip
+            audit_bundle_archive: output/audit_bundle.zip
+            delivery_artifact_count: 1
+            packaging_hygiene: clean
         - action: state.check
           exit_code: 0
         - action: status.show
@@ -398,6 +408,9 @@ cases:
         - output/intermediate/quality_panel.json
         - output/intermediate/quality_summary.md
         - output/intermediate/quality_panel.html
+        - output/report_bundle_manifest.json
+        - output/delivery_bundle.zip
+        - output/audit_bundle.zip
       artifact_statuses:
         - artifact_id: quality_panel
           status: valid
@@ -433,6 +446,12 @@ cases:
         - scope: workspace
           file: output/intermediate/quality_panel.html
           text: "Quality Panel"
+        - scope: workspace
+          file: output/report_bundle_manifest.json
+          text: "reader_facing_artifacts_only"
+        - scope: workspace
+          file: output/report_bundle_manifest.json
+          text: "audit_control_artifacts_only_not_reader_delivery"
       absent_text:
         - scope: workspace
           file: output/delivery/brief.md

src/multi_agent_brief/evaluation_cases/runner.py

@@ -41,6 +41,7 @@
     GUIDANCE_MANIFESTATION_REQUIRED_NON_GOALS,
     GUIDANCE_MANIFESTATION_RUNTIME_EFFECT,
 )
+from multi_agent_brief.product.bundle_projection import write_report_bundle_manifest
 from multi_agent_brief.product.quality_panel import (
     quality_panel_path,
     write_quality_panel,
@@ -503,6 +504,18 @@ def _dispatch_action(command: dict[str, Any], context: dict[str, Any]) -> dict[s
             result["quality_summary"] = data["quality_summary"]
         if "quality_panel_html" in data:
             result["quality_panel_html"] = data["quality_panel_html"]
+        if "report_bundle_manifest" in data:
+            result["report_bundle_manifest"] = data["report_bundle_manifest"]
+        if "delivery_bundle_archive" in data:
+            result["delivery_bundle_archive"] = data["delivery_bundle_archive"]
+        if "audit_bundle_archive" in data:
+            result["audit_bundle_archive"] = data["audit_bundle_archive"]
+        if "delivery_artifact_count" in data:
+            result["delivery_artifact_count"] = data["delivery_artifact_count"]
+        if "audit_artifact_count" in data:
+            result["audit_artifact_count"] = data["audit_artifact_count"]
+        if "packaging_hygiene" in data:
+            result["packaging_hygiene"] = data["packaging_hygiene"]
         if "suggested_next_command" in data:
             result["suggested_next_command"] = data["suggested_next_command"]
         return result
@@ -587,6 +600,8 @@ def _run_action(*, action: str, args: dict[str, Any], context: dict[str, Any]) -
         )
     if action == "quality.summarize":
         return _write_quality_projection_artifacts(workspace=_require_workspace(workspace))
+    if action == "packs.bundle":
+        return _write_bundle_projection_artifacts(workspace=_require_workspace(workspace))
     if action == "state.decide":
         return record_decision(
             workspace=_require_workspace(workspace),
@@ -724,6 +739,24 @@ def _write_quality_projection_artifacts(*, workspace: Path) -> dict[str, Any]:
     }
 
 
+def _write_bundle_projection_artifacts(*, workspace: Path) -> dict[str, Any]:
+    ws = workspace.expanduser().resolve()
+    manifest = write_report_bundle_manifest(workspace=ws, write_archives=True)
+    archives = manifest.get("bundle_archives") if isinstance(manifest.get("bundle_archives"), dict) else {}
+    delivery = archives.get("delivery") if isinstance(archives.get("delivery"), dict) else {}
+    audit = archives.get("audit") if isinstance(archives.get("audit"), dict) else {}
+    return {
+        "ok": True,
+        "report_bundle_manifest": manifest.get("manifest_path"),
+        "delivery_bundle_archive": delivery.get("path"),
+        "audit_bundle_archive": audit.get("path"),
+        "delivery_artifact_count": (manifest.get("delivery_bundle") or {}).get("artifact_count"),
+        "audit_artifact_count": (manifest.get("audit_bundle") or {}).get("artifact_count"),
+        "packaging_hygiene": (manifest.get("packaging_hygiene") or {}).get("status"),
+        "boundary": "bundle_projection_only_not_delivery_or_release_authority",
+    }
+
+
 def _seed_guidance_manifestation_report(*, workspace: Path, args: dict[str, Any]) -> dict[str, Any]:
     intermediate = workspace / "output" / "intermediate"
     manifest_path = intermediate / "runtime_manifest.json"

tests/test_evaluation_cases.py

@@ -87,13 +87,18 @@ def test_eval_cases_same_evidence_reader_quality_regression(capsys):
         "state.check",
         "status.show",
         "quality.summarize",
+        "packs.bundle",
         "state.check",
         "status.show",
     ]
     quality_action = case["actions"][2]
     assert quality_action["quality_panel"] == "output/intermediate/quality_panel.json"
     assert quality_action["quality_summary"] == "output/intermediate/quality_summary.md"
     assert quality_action["quality_panel_html"] == "output/intermediate/quality_panel.html"
+    bundle_action = case["actions"][3]
+    assert bundle_action["report_bundle_manifest"] == "output/report_bundle_manifest.json"
+    assert bundle_action["delivery_bundle_archive"] == "output/delivery_bundle.zip"
+    assert bundle_action["audit_bundle_archive"] == "output/audit_bundle.zip"
 
 
 def test_eval_cases_improvement_approved_case_materializes_snapshot(capsys):