diff --git a/.github/workflows/project-sync-safety.yml b/.github/workflows/project-sync-safety.yml
index 9362a5f..d7cd6cb 100644
--- a/.github/workflows/project-sync-safety.yml
+++ b/.github/workflows/project-sync-safety.yml
@@ -41,10 +41,10 @@ jobs:
 
     steps:
       - name: Check out repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v7
 
       - name: Set up Python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@v7
         with:
           python-version: "3.12"
           cache: pip
diff --git a/.github/workflows/quality.yml b/.github/workflows/quality.yml
index 78e9980..cc232db 100644
--- a/.github/workflows/quality.yml
+++ b/.github/workflows/quality.yml
@@ -29,11 +29,11 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y libegl1
       - run: python -m pip install --upgrade pip
-      - run: pip install -e ".[dev,web]"
+      - run: python -m pip install -e ".[dev,web]"
       - run: pre-commit validate-config
       - run: ruff check .
       - run: black --check --diff .
       - run: mypy src/docmergeforge
       - run: python scripts/check_docs_links.py
       - run: python scripts/check_repository_reference.py
-      - run: pytest --cov=docmergeforge --cov-report=term-missing
+      - run: python -m pytest --cov=docmergeforge --cov-report=term-missing
diff --git a/.github/workflows/regression.yml b/.github/workflows/regression.yml
index 0199990..545883d 100644
--- a/.github/workflows/regression.yml
+++ b/.github/workflows/regression.yml
@@ -28,7 +28,7 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y libegl1
       - run: python -m pip install --upgrade pip
-      - run: pip install -e ".[dev]"
+      - run: python -m pip install -e ".[dev,web]"
       - run: python scripts/generate_120_fixture.py fixtures/generated/sql-120
-      - run: pytest -m "regression or integration" tests/regression tests/integration
+      - run: python -m pytest -m "regression or integration" tests/regression tests/integration
       - run: docmergeforge validate --input fixtures/generated/sql-120 --parts 1-120
diff --git a/PROJECT_STATE.md b/PROJECT_STATE.md
index 60a82a7..162d757 100644
--- a/PROJECT_STATE.md
+++ b/PROJECT_STATE.md
@@ -5,25 +5,59 @@ This file is the compact continuation checkpoint for future development sessions
 ## Current checkpoint
 
 - Repository: `sanskarIN/DocMergeForge`
-- Branch: `main`
-- Version declared in `pyproject.toml`: `0.1.0`
-- Checkpoint immediately before this state-file commit: `b2379226f7681ffe64a2dcfa3b9d59c75006bf30`
-- Continuation base for the latest completed feature: `9775190f38e613e33f20aafc82b678a1ca3a233d`
-- Development status: pre-stable; do not claim `v1.0.0`, production certification, native mobile packaging, or completed signing/notarization from source changes alone.
+- Target branch: `main`
+- Active release-preparation branch: `release/2.8.5-prep`
+- Version declared in `pyproject.toml`: `2.8.5`
+- Release-preparation base: `aac5bf6d275991e21b68e15f5ad31f084fbc72e2`
+- Development/release status: `2.8.5` candidate preparation; package versioning does **not** by itself certify production readiness, native mobile packaging, signing/notarization, human accessibility acceptance, or external-office fidelity.
 
-## Latest completed continuation: guarded desktop project synchronization
+## Latest continuation: 2.8.5 release preparation
 
-The desktop application now exposes the existing review-first project synchronization model that was previously available through the CLI.
+The `2.8.5` preparation pass is intentionally evidence-first. It updates package/runtime version metadata and release-facing documentation without converting unobserved CI, packaging, device, accessibility, stress, signing, or fidelity gates into passing claims.
+
+### Release metadata
+
+The candidate version is synchronized across:
+
+- `pyproject.toml` — package version `2.8.5`;
+- `src/docmergeforge/__init__.py` — runtime `__version__ = "2.8.5"`;
+- `tests/unit/test_version_metadata.py` — package/runtime synchronization plus an explicit `2.8.5` candidate pin.
+
+The existing pre-alpha distribution classifier remains intentionally conservative. A numeric version change is not used as a substitute for release acceptance evidence.
+
+### Workflow-generation cleanup
+
+The focused Project Sync Safety workflow was the remaining workflow still using older GitHub Action majors. The release-preparation branch aligns it with the repository's maintained Node-24-era action generations:
+
+- `actions/checkout@v7`;
+- `actions/setup-python@v7`.
+
+This incorporates the substantive changes proposed independently by Dependabot PRs #3 and #4 into one reviewed release-preparation line. The PRs should be closed as superseded after the release branch is merged, not represented as separately required product work.
+
+### Recent-project synchronization status corrected
+
+The desktop convenience shortcut from **Recent Projects** into the guarded synchronization workflow is already implemented on current `main` and covered by integration tests. It is no longer future work.
+
+`ProjectSyncMainWindow` exposes both:
+
+- **Synchronize Project Sources** — browse for a project JSON;
+- **Synchronize Recent Project** — select from maintained recent-project history.
+
+Both routes converge on the same `_synchronize_project_path(...)` guarded workflow and therefore preserve the same preview, duplicate blocking, separate removal approval, exact-revision propagation, backup, stale-write, and metadata-only semantics.
+
+## Previous completed continuation: guarded desktop project synchronization
+
+The desktop application exposes the existing review-first project synchronization model that was previously available through the CLI.
 
 ### Maintained public desktop entry
 
-`pyproject.toml` now routes:
+`pyproject.toml` routes:
 
 ```text
 docmergeforge-gui = "docmergeforge.ui.desktop_entry:main"
 ```
 
-`src/docmergeforge/ui/desktop_entry.py` provides `ProjectSyncMainWindow`, an extension of the established desktop `MainWindow` that adds **Synchronize Project Sources** without duplicating project discovery/synchronization business logic.
+`src/docmergeforge/ui/desktop_entry.py` provides `ProjectSyncMainWindow`, an extension of the established desktop `MainWindow` that adds synchronization actions without duplicating project discovery/synchronization business logic.
 
 Normal installed desktop startup and normal packaged desktop startup both route through this synchronization-enabled entry.
 
@@ -31,7 +65,7 @@ Normal installed desktop startup and normal packaged desktop startup both route
 
 The maintained desktop flow is:
 
-1. select a saved project JSON;
+1. select a saved project JSON directly or through recent-project history;
 2. load the project and exact SHA-256 content revision from the same byte snapshot;
 3. call the shared `plan_project_sync(...)` planner;
 4. show a read-only preview containing current/proposed counts, additions, removals, reordering, duplicate parts, missing parts, and complete proposed order;
@@ -64,35 +98,42 @@ Desktop and CLI synchronization intentionally share these rules:
 
 The SHA-256 project revision mechanism is still an **optimistic stale-write guard**, not a universal cooperative cross-process lock. Do not represent simultaneous multi-writer editing as solved unless a separate coordinated locking/revision protocol is designed and accepted.
 
-## New/changed runtime paths in the latest continuation
+## Runtime and documentation paths associated with desktop synchronization
 
-### Added
+### Added in the desktop-sync continuation
 
 - `src/docmergeforge/ui/desktop_entry.py` — synchronization-enabled maintained desktop startup and workflow orchestration.
 - `src/docmergeforge/ui/project_sync_dialog.py` — accessible synchronization review dialog.
-- `tests/integration/test_project_sync_desktop.py` — offscreen Qt/workflow regression coverage for the new desktop path.
-- `docs/history/what_changed-through-2026-08-20-cross-platform.md` — verbatim archive of the previous top-level development record before this desktop-sync continuation.
+- `tests/integration/test_project_sync_desktop.py` — offscreen Qt/workflow regression coverage for the desktop path, including browse and recent-project routing.
+- `docs/history/what_changed-through-2026-08-20-cross-platform.md` — verbatim archive of the previous top-level development record before the desktop-sync continuation.
 
-### Changed
+### Changed across the synchronization/release-prep boundary
 
-- `pyproject.toml` — public GUI console-script target.
+- `pyproject.toml` — public GUI console-script target and `2.8.5` candidate package version.
+- `src/docmergeforge/__init__.py` — `2.8.5` runtime package version.
 - `src/docmergeforge/ui/packaged_entry.py` — packaged startup/smoke uses the synchronization-enabled window.
-- `src/docmergeforge/packaging/desktop.py` — packaging preflight requires the new desktop entry/dialog modules.
+- `src/docmergeforge/packaging/desktop.py` — packaging preflight requires the desktop entry/dialog modules.
+- `.github/workflows/project-sync-safety.yml` — focused safety matrix and maintained action majors.
 - `tests/unit/test_build_desktop.py` — packaging prerequisite coverage.
-- `tests/unit/test_version_metadata.py` — pins maintained CLI/GUI/web public entry points.
-- `README.md` — public desktop synchronization feature/safety description.
+- `tests/unit/test_version_metadata.py` — pins maintained CLI/GUI/web public entry points and the `2.8.5` candidate version.
+- `README.md` — public desktop synchronization feature/safety description and release-preparation status.
 - `docs/desktop-guide.md` — operator workflow.
 - `docs/project-sync.md` — shared desktop/CLI synchronization contract.
 - `docs/source-code-reference.md` — runtime responsibility map.
 - `docs/test-suite-reference.md` — test ownership/evidence map.
-- `docs/repository-reference-cross-platform.md` — tracked-path coverage for the new files/archive.
-- `what_changed.md` — current continuation record only; the preceding complete record is archived under `docs/history/`.
+- `docs/repository-reference-cross-platform.md` — tracked-path coverage for the cross-platform/desktop additions.
+- `docs/release-process.md` — versioning/release evidence policy.
+- `docs/release-evidence.md` — evidence ledger; candidate evidence must use exact run/checkpoint IDs.
+- `CHANGELOG.md` — release-preparation record.
+- `what_changed.md` — active continuation record.
 
 ## Regression coverage added/expanded
 
 `tests/integration/test_project_sync_desktop.py` protects:
 
-- presence/accessibility of the desktop synchronization action;
+- presence/accessibility of both desktop synchronization actions;
+- browse-project routing;
+- recent-project routing;
 - accessible complete preview content;
 - disabled apply for ambiguous duplicate parts;
 - exact revision propagation into the shared apply path;
@@ -104,8 +145,8 @@ The SHA-256 project revision mechanism is still an **optimistic stale-write guar
 Additional related coverage:
 
 - `tests/unit/test_build_desktop.py` requires the base window, desktop entry, sync dialog, and packaged entry in build-root preflight;
-- `tests/unit/test_version_metadata.py` pins `docmergeforge`, `docmergeforge-gui`, and `docmergeforge-web` entry targets;
-- `tests/integration/test_packaged_entry_smoke.py` now reaches the synchronization-enabled packaged window before the existing real temporary PDF/DOCX publication smoke.
+- `tests/unit/test_version_metadata.py` pins `docmergeforge`, `docmergeforge-gui`, and `docmergeforge-web` entry targets and candidate version metadata;
+- `tests/integration/test_packaged_entry_smoke.py` reaches the synchronization-enabled packaged window before the existing real temporary PDF/DOCX publication smoke.
 
 Committed test source is implementation evidence only until execution is observed.
 
@@ -146,25 +187,15 @@ The tracked-file documentation checker reads the maintained reference corpus:
 - `docs/repository-reference.md`;
 - `docs/repository-reference-cross-platform.md`.
 
-The latest continuation added exact backticked references for every new tracked runtime/test/history path before or alongside the corresponding file becoming part of the maintained checkpoint.
-
-The checker is configured in Quality and pre-commit. Configuration is not the same as an observed passing current-head execution.
+The checker is configured in Quality and pre-commit. Configuration is not the same as an observed passing candidate-head execution.
 
 ## Verification boundary
 
 Do not infer a green build merely from commits being present.
 
-During the latest continuation:
-
-- focused GitHub commit/tree/file inspection was used to keep changes scoped;
-- packaging/source/test/documentation dependencies were cross-checked against the repository;
-- an unintended temporary Ruff-rule-set expansion introduced while changing the GUI entry was immediately reverted in the next focused commit;
-- the maintained Ruff rule set remains `E/F/I/B/UP/SIM/C4`;
-- a raw GitHub archive/checkout could not be obtained in the execution environment, so local quality/test execution was not available;
-- repository workflow definitions remain configured for Quality/Build Smoke, but no fresh passing current-head execution is claimed without observed run evidence.
-
-Until observed for the current head, no fresh pass is claimed for:
+For the `2.8.5` candidate, no fresh pass is claimed until the exact release-preparation head is observed for the relevant gates. Required source evidence includes:
 
+- pre-commit configuration validation;
 - Ruff;
 - Black check;
 - strict mypy;
@@ -174,31 +205,42 @@ Until observed for the current head, no fresh pass is claimed for:
 - Quality workflow matrix;
 - 120-Part Regression;
 - Build Smoke;
-- Package Desktop / Onefile Acceptance;
-- Security/CodeQL;
+- Project Sync Safety matrix;
+- Security/CodeQL.
+
+Packaging/release evidence remains separate:
+
+- Package Desktop / Onefile Acceptance on Windows, macOS, and Linux;
+- downloaded-artifact verification;
 - representative Android/iOS/iPadOS/ChromeOS/manual browser acceptance;
-- human desktop accessibility/clean-machine acceptance.
+- human desktop accessibility/clean-machine acceptance;
+- external-office fidelity acceptance;
+- measured stress at the workload class actually claimed;
+- Windows signing;
+- macOS signing/notarization;
+- final distribution/installer acceptance where applicable.
 
-External-office, measured stress, signing/notarization, and other release gates remain independent.
+Older recorded passing runs remain historical evidence for their exact checkpoints and must not be relabeled as `2.8.5` candidate evidence.
 
-## Repository administration state observed during this continuation
+## Repository administration state
 
-GitHub branch metadata reported `main` as not protected, with required status checks disabled at the repository-rules layer at the inspected checkpoint.
+At the release-preparation base, GitHub branch metadata reported `main` as not protected, with required status checks disabled at the repository-rules layer.
 
 This is an administrative governance state, not an application correctness failure. If enforced review/CI policy on `main` is desired, configure branch protection/rulesets through repository administration with the intended required checks. Do not claim protection is enabled until repository metadata confirms it.
 
 ## Recommended next development work
 
-1. Observe a current-head Quality run; fix any lint/format/type/test/link/reference failure without weakening maintained rules.
-2. Review current Build Smoke and packaged-app results specifically for the synchronization-enabled desktop entry on Windows, macOS, and Linux.
-3. If a convenience shortcut from **Recent Projects** into synchronization is added, preserve the same preview, duplicate blocking, separate removal approval, exact revision, and backup semantics; do not bypass them.
-4. Keep synchronization domain rules centralized in `project.sync` rather than forking CLI and desktop business logic.
+1. Open/review the `release/2.8.5-prep` pull request and observe Quality plus the focused Project Sync Safety checks for the exact candidate head.
+2. Fix any Ruff/Black/mypy/docs/reference/pytest/CI failure without weakening maintained checks.
+3. Review current 120-Part Regression, Build Smoke, Security/CodeQL, Package Desktop, and Onefile Acceptance evidence for the exact candidate commit.
+4. Keep synchronization domain rules centralized in `project.sync`; both browse and recent-project desktop actions must continue to converge on the shared guarded workflow.
 5. Perform representative manual browser/device acceptance for the responsive cross-platform client.
 6. If Internet/untrusted-network hosting is intentionally supported later, define and acceptance-test an explicit HTTPS reverse-proxy/authentication/body-limit/timeout/concurrency/host-hardening deployment profile.
 7. If simultaneous multi-writer project editing becomes a supported requirement, design a separate coordinated lock/revision protocol rather than relabeling the optimistic revision guard.
 8. Continue independent release-gate work for native-office fidelity, measured multi-gigabyte stress, human accessibility, clean-machine packaged applications, Windows signing, and macOS signing/notarization.
-9. Keep README, project-sync/desktop/source/test references, repository-reference corpus, `what_changed.md`, and this checkpoint synchronized whenever the boundary changes.
+9. Native Android APK/AAB and native iOS IPA delivery remain separate implementation tracks; browser support must not be relabeled as native packaging.
+10. Keep README, changelog, release-process/evidence docs, project-sync/desktop/source/test references, repository-reference corpus, `what_changed.md`, and this checkpoint synchronized whenever the release boundary changes.
 
 ## Continuation rule
 
-Future sessions should inspect the actual current `main` head, read this file plus `what_changed.md`, and consult the repository/source/test/automation/configuration references before modifying a subsystem. Continue from repository evidence instead of re-opening completed work, and never turn configured automation or committed tests into claimed passing evidence without an observed run.
+Future sessions should inspect the actual current `main` and any active release branch, read this file plus `what_changed.md`, and consult the repository/source/test/automation/configuration/release references before modifying a subsystem. Continue from repository evidence instead of re-opening completed work, and never turn configured automation or committed tests into claimed passing evidence without an observed run for the exact checkpoint.
diff --git a/docs/release-process.md b/docs/release-process.md
index 8893889..e0b6134 100644
--- a/docs/release-process.md
+++ b/docs/release-process.md
@@ -1,8 +1,8 @@
 # Release Process
 
-This document defines the evidence required to move DocMergeForge from a development checkpoint to a release candidate and, eventually, a stable `v1.0.0` release.
+This document defines the evidence required to move DocMergeForge from a development checkpoint to a release candidate and, eventually, an intentionally production-accepted release.
 
-A release is not complete merely because source tests pass or PyInstaller creates an archive.
+A release is not complete merely because source tests pass, a version number changes, or PyInstaller creates an archive.
 
 ## Release evidence levels
 
@@ -17,7 +17,9 @@ Documentation, changelogs, and releases must not collapse these levels into a si
 
 ## Versioning
 
-The project follows semantic-versioning intent. The package remains pre-stable (`0.x`) while the release matrix is being completed. `1.0.0` is reserved for the first intentionally accepted stable public contract.
+The current release-preparation target is `2.8.5`. DocMergeForge uses semantic-version-shaped package identifiers, but the repository's evidence model remains authoritative for readiness: a major version greater than zero or one does **not** by itself mean that human/production acceptance, signing/notarization, native mobile packaging, external-office fidelity, or claimed-scale stress gates have passed.
+
+The distribution classifier and release documentation must remain conservative until the corresponding evidence is accepted. For `2.8.5`, treat package/runtime metadata as a candidate identifier first; promote/tag/publish only after the intended release gates for that distribution claim are green and reviewed.
 
 ## 1. Freeze release scope
 
@@ -29,13 +31,16 @@ Review/update as appropriate:
 
 ```text
 pyproject.toml
+src/docmergeforge/__init__.py
+tests/unit/test_version_metadata.py
 CHANGELOG.md
 README.md
 docs/
+PROJECT_STATE.md
 what_changed.md
 ```
 
-Every documented command and support claim must match the candidate implementation.
+Every documented command and support claim must match the candidate implementation. Package metadata and runtime `__version__` must agree, and the release candidate should have an explicit regression pin so accidental version drift fails before publication.
 
 ## 3. Source quality and documentation gate
 
@@ -47,6 +52,7 @@ ruff check .
 black --check --diff .
 mypy src/docmergeforge
 python scripts/check_docs_links.py
+python scripts/check_repository_reference.py
 pytest --cov=docmergeforge --cov-report=term-missing
 ```
 
@@ -111,7 +117,7 @@ Run:        32033135355
 Checkpoint: 59dc14bbf1d4301177e475ac350694bdd9d90ada
 ```
 
-All Windows/macOS/Ubuntu build-host and fresh-runner jobs passed.
+All Windows/macOS/Ubuntu build-host and fresh-runner jobs passed for that historical checkpoint. Re-run/review candidate-appropriate packaging evidence before attributing it to `2.8.5`.
 
 ## 13. Optional onefile gate
 
@@ -124,7 +130,7 @@ Run:        32033541414
 Checkpoint: dc624e23d07e0ce94ef345245630d153ee60091a
 ```
 
-All Windows/macOS/Ubuntu build-host and fresh-runner jobs passed.
+All Windows/macOS/Ubuntu build-host and fresh-runner jobs passed for that historical checkpoint. Do not relabel it as `2.8.5` evidence unless the exact candidate commit is the checkpoint being verified.
 
 ## 14. Build provenance and SBOM gate
 
@@ -149,7 +155,7 @@ Before a production support claim, use representative clean end-user machines/VM
 
 ## 16. Platform signing/notarization gate
 
-Current artifacts are explicitly unsigned development builds.
+Current artifacts are explicitly unsigned development builds unless final-stage evidence says otherwise.
 
 ### Windows
 
@@ -185,7 +191,7 @@ Release notes should state supported platforms/architectures/build modes, worklo
 
 A `v*` tag can trigger packaging, but the tag itself is not acceptance and does not make current artifacts signed.
 
-Tag only the chosen reviewed commit after required gates for the release claim are green/accepted.
+Tag only the chosen reviewed commit after required gates for the release claim are green/accepted. For the current cycle, do not create or describe `v2.8.5` as accepted merely because package metadata says `2.8.5`.
 
 ## 21. Post-release verification
 
@@ -199,9 +205,11 @@ After publishing through the real user-facing channel:
 6. confirm no private/debug files were uploaded;
 7. retain run IDs, provenance/SBOM/checksums, trust evidence, and human acceptance records.
 
-## Stable `v1.0.0` gate
+## Production-acceptance gate
+
+Do not claim a production-accepted release until required areas for the intended support statement are intentionally accepted, including core merge correctness, transaction/recovery safety, representative large/stress workloads, real-world fidelity, human accessibility, downloaded-artifact and human clean-machine package acceptance, production signing/notarization where distributed, and complete documentation/support/security processes.
 
-Do not claim `v1.0.0` until required areas for the intended support statement are intentionally accepted, including core merge correctness, transaction/recovery safety, representative large/stress workloads, real-world fidelity, human accessibility, downloaded-artifact and human clean-machine package acceptance, production signing/notarization where distributed, and complete documentation/support/security processes.
+This gate applies to `2.8.5` regardless of its numeric major version.
 
 ## Release evidence template
 
@@ -210,6 +218,7 @@ Version/tag:
 Commit SHA:
 Date:
 Quality run:
+Project Sync Safety run:
 120-Part Regression run:
 Build Smoke run:
 Security run:
diff --git a/pyproject.toml b/pyproject.toml
index 263c34d..2546b23 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "docmergeforge"
-version = "0.1.0"
+version = "2.8.5"
 description = "Local-first, validation-driven PDF and DOCX merger for large multi-part publications."
 readme = "README.md"
 requires-python = ">=3.12"
diff --git a/scripts/check_docx_fidelity_acceptance.py b/scripts/check_docx_fidelity_acceptance.py
index ab3e0c3..6763c9b 100644
--- a/scripts/check_docx_fidelity_acceptance.py
+++ b/scripts/check_docx_fidelity_acceptance.py
@@ -13,9 +13,7 @@ def build_fixture(path: Path) -> None:
     document = Document()
     document.core_properties.title = "DocMergeForge Fidelity Acceptance"
     document.add_heading("Fidelity Acceptance", level=1)
-    document.add_paragraph(
-        "Representative smoke content with bold, italic, and list formatting."
-    )
+    document.add_paragraph("Representative smoke content with bold, italic, and list formatting.")
     formatted = document.add_paragraph()
     formatted.add_run("Bold text").bold = True
     formatted.add_run(" and ")
diff --git a/src/docmergeforge/__init__.py b/src/docmergeforge/__init__.py
index 48f7c91..c986183 100644
--- a/src/docmergeforge/__init__.py
+++ b/src/docmergeforge/__init__.py
@@ -1,3 +1,3 @@
 """DocMergeForge package."""
 
-__version__ = "0.1.0"
+__version__ = "2.8.5"
diff --git a/src/docmergeforge/cli/main.py b/src/docmergeforge/cli/main.py
index 65b6c8f..480a58d 100644
--- a/src/docmergeforge/cli/main.py
+++ b/src/docmergeforge/cli/main.py
@@ -278,9 +278,7 @@ def _run_direct_merge(args: argparse.Namespace) -> int:
                         "ready": False,
                         "missing": validation_result.missing_parts,
                         "duplicates": validation_result.duplicate_parts,
-                        "diagnostics": [
-                            item.to_dict() for item in validation_result.diagnostics
-                        ],
+                        "diagnostics": [item.to_dict() for item in validation_result.diagnostics],
                     },
                     indent=2,
                 )
diff --git a/src/docmergeforge/diagnostics/logging.py b/src/docmergeforge/diagnostics/logging.py
index 5acbb9d..dbbe678 100644
--- a/src/docmergeforge/diagnostics/logging.py
+++ b/src/docmergeforge/diagnostics/logging.py
@@ -15,9 +15,7 @@
     r"(?P=quote)?"
 )
 _BEARER_PATTERN = re.compile(r"(?i)\bBearer\s+[A-Za-z0-9._~+/=-]+")
-_AUTH_HEADER_PATTERN = re.compile(
-    r"(?i)\bAuthorization\s*:\s*(?:Basic|Bearer)\s+[^\s,;]+"
-)
+_AUTH_HEADER_PATTERN = re.compile(r"(?i)\bAuthorization\s*:\s*(?:Basic|Bearer)\s+[^\s,;]+")
 _API_KEY_HEADER_PATTERN = re.compile(r"(?i)\b(?:X-)?Api-Key\s*:\s*[^\s,;]+")
 
 
diff --git a/src/docmergeforge/discovery/scanner.py b/src/docmergeforge/discovery/scanner.py
index e5f099f..51abc7a 100644
--- a/src/docmergeforge/discovery/scanner.py
+++ b/src/docmergeforge/discovery/scanner.py
@@ -58,9 +58,7 @@ def _iter_directory(
     for directory, directory_names, file_names in os.walk(root, followlinks=False):
         directory_path = Path(directory)
         directory_names[:] = [
-            name
-            for name in directory_names
-            if not _is_excluded(directory_path / name, excluded)
+            name for name in directory_names if not _is_excluded(directory_path / name, excluded)
         ]
         for name in file_names:
             path = directory_path / name
diff --git a/src/docmergeforge/docx/fidelity_acceptance.py b/src/docmergeforge/docx/fidelity_acceptance.py
index 3ed5140..d8df50a 100644
--- a/src/docmergeforge/docx/fidelity_acceptance.py
+++ b/src/docmergeforge/docx/fidelity_acceptance.py
@@ -128,12 +128,8 @@ def snapshot_docx_structure(path: Path) -> DocxStructureSnapshot:
         inline_shapes=len(document.inline_shapes),
         sections=len(document.sections),
         headings=headings,
-        header_paragraphs=sum(
-            len(section.header.paragraphs) for section in document.sections
-        ),
-        footer_paragraphs=sum(
-            len(section.footer.paragraphs) for section in document.sections
-        ),
+        header_paragraphs=sum(len(section.header.paragraphs) for section in document.sections),
+        footer_paragraphs=sum(len(section.footer.paragraphs) for section in document.sections),
         header_tables=sum(len(section.header.tables) for section in document.sections),
         footer_tables=sum(len(section.footer.tables) for section in document.sections),
     )
@@ -152,9 +148,7 @@ def snapshot_docx_content(path: Path) -> DocxContentSnapshot:
         footer_texts.extend(_table_texts(section.footer.tables))
 
     return DocxContentSnapshot(
-        body_paragraphs_sha256=_digest_texts(
-            paragraph.text for paragraph in document.paragraphs
-        ),
+        body_paragraphs_sha256=_digest_texts(paragraph.text for paragraph in document.paragraphs),
         tables_sha256=_digest_texts(_table_texts(document.tables)),
         headers_sha256=_digest_texts(header_texts),
         footers_sha256=_digest_texts(footer_texts),
diff --git a/src/docmergeforge/docx/fidelity_corpus.py b/src/docmergeforge/docx/fidelity_corpus.py
index 2ed4065..1b12f6c 100644
--- a/src/docmergeforge/docx/fidelity_corpus.py
+++ b/src/docmergeforge/docx/fidelity_corpus.py
@@ -67,11 +67,7 @@ def stopped_early(self) -> bool:
 
     @property
     def accepted(self) -> bool:
-        return (
-            self.discovered_count > 0
-            and not self.stopped_early
-            and self.failed_count == 0
-        )
+        return self.discovered_count > 0 and not self.stopped_early and self.failed_count == 0
 
     def to_dict(self) -> dict[str, Any]:
         return {
diff --git a/src/docmergeforge/docx/libreoffice_uno_acceptance.py b/src/docmergeforge/docx/libreoffice_uno_acceptance.py
index 8940693..6c5507f 100644
--- a/src/docmergeforge/docx/libreoffice_uno_acceptance.py
+++ b/src/docmergeforge/docx/libreoffice_uno_acceptance.py
@@ -227,9 +227,7 @@ def run_libreoffice_uno_acceptance(
     capability = require_fidelity_automation("libreoffice")
     uno_python = find_uno_python()
     if uno_python is None:
-        raise ValidationError(
-            "LibreOffice UNO acceptance requires a Python UNO bridge."
-        )
+        raise ValidationError("LibreOffice UNO acceptance requires a Python UNO bridge.")
     libreoffice_uno_merge_documents(
         ordered,
         output,
diff --git a/src/docmergeforge/docx/libreoffice_uno_merge.py b/src/docmergeforge/docx/libreoffice_uno_merge.py
index 470748f..1eb165d 100644
--- a/src/docmergeforge/docx/libreoffice_uno_merge.py
+++ b/src/docmergeforge/docx/libreoffice_uno_merge.py
@@ -20,7 +20,7 @@
 )
 from docmergeforge.utilities.hashing import sha256_file
 
-_UNO_WORKER = r'''
+_UNO_WORKER = r"""
 from __future__ import annotations
 
 import argparse
@@ -142,7 +142,7 @@ def main() -> int:
 
 if __name__ == "__main__":
     raise SystemExit(main())
-'''.strip()
+""".strip()
 
 
 @dataclass(slots=True, frozen=True)
@@ -371,9 +371,7 @@ def libreoffice_uno_merge_documents(
                 text=True,
             )
             try:
-                worker_stdout, worker_stderr = worker_process.communicate(
-                    timeout=timeout_seconds
-                )
+                worker_stdout, worker_stderr = worker_process.communicate(timeout=timeout_seconds)
             except subprocess.TimeoutExpired as exc:
                 worker_process.kill()
                 worker_stdout, worker_stderr = worker_process.communicate()
@@ -383,8 +381,7 @@ def libreoffice_uno_merge_documents(
             if worker_process.returncode != 0:
                 detail = worker_stderr.strip() or worker_stdout.strip()
                 raise ValidationError(
-                    "LibreOffice UNO merge worker failed"
-                    + (f": {detail}" if detail else ".")
+                    "LibreOffice UNO merge worker failed" + (f": {detail}" if detail else ".")
                 )
         finally:
             if worker_process is not None and worker_process.poll() is None:
diff --git a/src/docmergeforge/docx/section_evidence.py b/src/docmergeforge/docx/section_evidence.py
index f0b3022..d4e0bad 100644
--- a/src/docmergeforge/docx/section_evidence.py
+++ b/src/docmergeforge/docx/section_evidence.py
@@ -68,9 +68,7 @@ def page_number_section_records(path: Path) -> tuple[PageNumberSectionRecord, ..
             try:
                 document_xml = archive.read(_DOCUMENT_XML)
             except KeyError as exc:
-                raise ValidationError(
-                    f"DOCX package is missing {_DOCUMENT_XML}: {path}"
-                ) from exc
+                raise ValidationError(f"DOCX package is missing {_DOCUMENT_XML}: {path}") from exc
     except zipfile.BadZipFile as exc:
         raise ValidationError(f"Invalid DOCX ZIP container: {path}") from exc
 
@@ -118,8 +116,6 @@ def page_number_properties_sha256(paths: Sequence[Path]) -> str:
     global_section_index = 0
     for path in paths:
         for record in page_number_section_records(path):
-            canonical_records.append(
-                f"section={global_section_index}|{record.canonical()}"
-            )
+            canonical_records.append(f"section={global_section_index}|{record.canonical()}")
             global_section_index += 1
     return _digest_records(canonical_records)
diff --git a/src/docmergeforge/docx/word_merge_acceptance.py b/src/docmergeforge/docx/word_merge_acceptance.py
index 84e8523..e7f7ca9 100644
--- a/src/docmergeforge/docx/word_merge_acceptance.py
+++ b/src/docmergeforge/docx/word_merge_acceptance.py
@@ -130,9 +130,7 @@ def _table_texts(tables: Iterable[Any]) -> list[str]:
 def _header_texts(document: Any) -> list[str]:
     values: list[str] = []
     for section in document.sections:
-        values.extend(
-            paragraph.text for paragraph in section.header.paragraphs if paragraph.text
-        )
+        values.extend(paragraph.text for paragraph in section.header.paragraphs if paragraph.text)
         values.extend(_table_texts(section.header.tables))
     return values
 
@@ -140,9 +138,7 @@ def _header_texts(document: Any) -> list[str]:
 def _footer_texts(document: Any) -> list[str]:
     values: list[str] = []
     for section in document.sections:
-        values.extend(
-            paragraph.text for paragraph in section.footer.paragraphs if paragraph.text
-        )
+        values.extend(paragraph.text for paragraph in section.footer.paragraphs if paragraph.text)
         values.extend(_table_texts(section.footer.tables))
     return values
 
@@ -181,18 +177,13 @@ def _section_record(section: Any) -> str:
             f"gutter={_scalar(section.gutter)}",
             f"header_distance={_scalar(section.header_distance)}",
             f"footer_distance={_scalar(section.footer_distance)}",
-            "different_first_page="
-            f"{int(bool(section.different_first_page_header_footer))}",
+            "different_first_page=" f"{int(bool(section.different_first_page_header_footer))}",
             f"header_linked={int(bool(section.header.is_linked_to_previous))}",
-            "first_header_linked="
-            f"{int(bool(section.first_page_header.is_linked_to_previous))}",
-            "even_header_linked="
-            f"{int(bool(section.even_page_header.is_linked_to_previous))}",
+            "first_header_linked=" f"{int(bool(section.first_page_header.is_linked_to_previous))}",
+            "even_header_linked=" f"{int(bool(section.even_page_header.is_linked_to_previous))}",
             f"footer_linked={int(bool(section.footer.is_linked_to_previous))}",
-            "first_footer_linked="
-            f"{int(bool(section.first_page_footer.is_linked_to_previous))}",
-            "even_footer_linked="
-            f"{int(bool(section.even_page_footer.is_linked_to_previous))}",
+            "first_footer_linked=" f"{int(bool(section.first_page_footer.is_linked_to_previous))}",
+            "even_footer_linked=" f"{int(bool(section.even_page_footer.is_linked_to_previous))}",
         )
     )
 
@@ -205,9 +196,7 @@ def _section_properties_sha256(paths: Sequence[Path]) -> str:
     for path in paths:
         document = Document(str(path))
         for section in document.sections:
-            records.append(
-                f"section={global_section_index}|{_section_record(section)}"
-            )
+            records.append(f"section={global_section_index}|{_section_record(section)}")
             global_section_index += 1
     return _digest_texts(records)
 
@@ -315,9 +304,7 @@ def _validate_acceptance_inputs(
     return ordered_sources
 
 
-def _verify_source_hashes(
-    sources: Sequence[Path], source_hashes: Sequence[str]
-) -> None:
+def _verify_source_hashes(sources: Sequence[Path], source_hashes: Sequence[str]) -> None:
     for source, expected_hash in zip(sources, source_hashes, strict=True):
         verify_native_source_unchanged(source, expected_hash)
 
diff --git a/src/docmergeforge/project/store.py b/src/docmergeforge/project/store.py
index caf36f6..7800351 100644
--- a/src/docmergeforge/project/store.py
+++ b/src/docmergeforge/project/store.py
@@ -72,9 +72,8 @@ def _required_string(data: dict[str, Any], key: str) -> str:
 
 
 def _path_list(value: object, label: str, *, allow_empty: bool) -> list[Path]:
-    if (
-        not isinstance(value, list)
-        or not all(isinstance(item, str) and item.strip() for item in value)
+    if not isinstance(value, list) or not all(
+        isinstance(item, str) and item.strip() for item in value
     ):
         raise ValueError(f"Project field '{label}' must be a JSON array of non-empty path strings.")
     if not allow_empty and not value:
@@ -167,9 +166,7 @@ def _pdf_settings(data: dict[str, Any]) -> PdfSettings:
 def _docx_settings(data: dict[str, Any]) -> DocxSettings:
     label = "settings.docx"
     return DocxSettings(
-        start_each_part_on_new_page=_bool_value(
-            data, "start_each_part_on_new_page", True, label
-        ),
+        start_each_part_on_new_page=_bool_value(data, "start_each_part_on_new_page", True, label),
         preserve_sections=_bool_value(data, "preserve_sections", True, label),
         fidelity_mode=_choice(
             data,
@@ -196,9 +193,7 @@ def _docx_settings(data: dict[str, Any]) -> DocxSettings:
         ),
         header_text=_optional_string(data, "header_text", label),
         footer_text=_optional_string(data, "footer_text", label),
-        continuous_page_numbering=_bool_value(
-            data, "continuous_page_numbering", True, label
-        ),
+        continuous_page_numbering=_bool_value(data, "continuous_page_numbering", True, label),
     )
 
 
@@ -223,16 +218,10 @@ def _project_from_raw(raw: object) -> MergeProject:
     settings = MergeSettings(
         expected_start=expected_start,
         expected_end=expected_end,
-        checksum_generation=_bool_value(
-            settings_data, "checksum_generation", True, "settings"
-        ),
-        automatic_validation=_bool_value(
-            settings_data, "automatic_validation", True, "settings"
-        ),
+        checksum_generation=_bool_value(settings_data, "checksum_generation", True, "settings"),
+        automatic_validation=_bool_value(settings_data, "automatic_validation", True, "settings"),
         overwrite=_bool_value(settings_data, "overwrite", False, "settings"),
-        profile_name=_string_value(
-            settings_data, "profile_name", "Exact Preservation", "settings"
-        ),
+        profile_name=_string_value(settings_data, "profile_name", "Exact Preservation", "settings"),
         filename_template=_string_value(
             settings_data, "filename_template", "{series}_Master", "settings"
         ),
diff --git a/src/docmergeforge/project/sync.py b/src/docmergeforge/project/sync.py
index f9683d2..4e58fb1 100644
--- a/src/docmergeforge/project/sync.py
+++ b/src/docmergeforge/project/sync.py
@@ -78,7 +78,9 @@ def to_dict(self) -> dict[str, object]:
         }
 
 
-def _eligible_documents(project: MergeProject, discovered: list[InputDocument]) -> list[InputDocument]:
+def _eligible_documents(
+    project: MergeProject, discovered: list[InputDocument]
+) -> list[InputDocument]:
     start = project.settings.expected_start
     end = project.settings.expected_end
     unique: list[InputDocument] = []
@@ -127,9 +129,7 @@ def _missing_parts(
     end: int,
 ) -> tuple[int, ...]:
     found = {
-        item.part.number
-        for item in documents
-        if item.kind == kind and item.part.number is not None
+        item.part.number for item in documents if item.kind == kind and item.part.number is not None
     }
     if not found:
         return ()
@@ -157,9 +157,7 @@ def plan_project_sync(
     proposed_set = set(proposed_keys)
 
     added = tuple(
-        path
-        for path, key in zip(proposed, proposed_keys, strict=True)
-        if key not in current_set
+        path for path, key in zip(proposed, proposed_keys, strict=True) if key not in current_set
     )
     removed = tuple(
         path for path, key in zip(current, current_keys, strict=True) if key not in proposed_set
diff --git a/src/docmergeforge/ui/desktop_entry.py b/src/docmergeforge/ui/desktop_entry.py
index dfea376..085d0af 100644
--- a/src/docmergeforge/ui/desktop_entry.py
+++ b/src/docmergeforge/ui/desktop_entry.py
@@ -32,7 +32,8 @@ def __init__(self) -> None:
         self.sync_project_button = QPushButton("Synchronize Project Sources")
         self.sync_project_button.setAccessibleName("Synchronize project sources")
         self.sync_project_button.setAccessibleDescription(
-            "Browse for a saved project, preview source-selection changes, and optionally apply them."
+            "Browse for a saved project, preview source-selection changes, "
+            "and optionally apply them."
         )
         self.sync_project_button.setMinimumHeight(58)
         self.sync_project_button.clicked.connect(self._synchronize_project)
diff --git a/src/docmergeforge/ui/main.py b/src/docmergeforge/ui/main.py
index 19cbbdb..4cd8401 100644
--- a/src/docmergeforge/ui/main.py
+++ b/src/docmergeforge/ui/main.py
@@ -254,9 +254,7 @@ def _confirm_project_order(self, project: MergeProject, *, checkpoint: bool = Tr
         if order_dialog.exec() != int(order_dialog.DialogCode.Accepted):
             return False
         project.selected_files = order_dialog.ordered_paths()
-        if checkpoint and not self._checkpoint_project(project, "ordering"):
-            return False
-        return True
+        return not checkpoint or self._checkpoint_project(project, "ordering")
 
     def _new_project(self, initial_source: Path | None = None) -> None:
         dialog = ProjectSetupDialog(initial_source)
diff --git a/src/docmergeforge/ui/project_sync_dialog.py b/src/docmergeforge/ui/project_sync_dialog.py
index f19b734..20840a2 100644
--- a/src/docmergeforge/ui/project_sync_dialog.py
+++ b/src/docmergeforge/ui/project_sync_dialog.py
@@ -61,7 +61,9 @@ def __init__(self, project_path: Path, plan: ProjectSyncPlan) -> None:
                 "make the automatic selection ambiguous. Resolve the duplicates and preview again."
             )
         elif not plan.changed:
-            guidance = "The saved selected-file list already matches the current automatic proposal."
+            guidance = (
+                "The saved selected-file list already matches the current automatic proposal."
+            )
         elif plan.removed:
             guidance = (
                 "Review the removals carefully. Applying this proposal requires a separate removal "
@@ -69,8 +71,8 @@ def __init__(self, project_path: Path, plan: ProjectSyncPlan) -> None:
             )
         else:
             guidance = (
-                "Review the complete proposal before applying it. A versioned backup of the project "
-                "JSON will be created before the guarded update."
+                "Review the complete proposal before applying it. A versioned backup of the "
+                "project JSON will be created before the guarded update."
             )
 
         self.guidance = QLabel(guidance)
diff --git a/src/docmergeforge/validation/service.py b/src/docmergeforge/validation/service.py
index b9e8923..6484ca2 100644
--- a/src/docmergeforge/validation/service.py
+++ b/src/docmergeforge/validation/service.py
@@ -46,9 +46,7 @@ def validate_part_set(
                 )
             )
         if kind == DocumentKind.PDF and item.encrypted and is_merge_input:
-            part_label = (
-                f"Part {item.part.number}" if item.part.number is not None else "Selected"
-            )
+            part_label = f"Part {item.part.number}" if item.part.number is not None else "Selected"
             if allow_encrypted_pdf:
                 diagnostics.append(
                     Diagnostic(
diff --git a/src/docmergeforge/web/app.py b/src/docmergeforge/web/app.py
index 1d534d5..17e5ab5 100644
--- a/src/docmergeforge/web/app.py
+++ b/src/docmergeforge/web/app.py
@@ -45,12 +45,25 @@
     input, button { box-sizing: border-box; width: 100%; min-height: 48px; border-radius: 12px; }
     input { border: 1px solid #46547a; background: #0f1629; color: #eef2ff; padding: 10px 12px; }
     input[type=file] { padding: 9px; }
-    button { margin-top: 20px; border: 0; background: #6d7cff; color: white; font-weight: 800; cursor: pointer; }
+    button {
+      margin-top: 20px;
+      border: 0;
+      background: #6d7cff;
+      color: white;
+      font-weight: 800;
+      cursor: pointer;
+    }
     button:disabled { opacity: .55; cursor: progress; }
     .muted { color: #b7c0d9; }
     .status { min-height: 28px; margin-top: 16px; font-weight: 650; }
     .grid { display: grid; gap: 14px; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); }
-    .pill { display: inline-block; border: 1px solid #46547a; border-radius: 999px; padding: 6px 10px; margin: 4px 4px 0 0; }
+    .pill {
+      display: inline-block;
+      border: 1px solid #46547a;
+      border-radius: 999px;
+      padding: 6px 10px;
+      margin: 4px 4px 0 0;
+    }
     a { color: #aeb8ff; }
     @media (prefers-color-scheme: light) {
       body { background: #f4f6fb; color: #182035; }
@@ -64,12 +77,25 @@
 <main>
   <section class="card">
     <h1>DocMergeForge</h1>
-    <p>Merge PDF or DOCX parts from Windows, macOS, Linux, Android, iPhone/iPad, ChromeOS, or any modern browser.</p>
-    <p class="muted">Files are processed by the DocMergeForge Python host you connected to. The default server binds only to this computer; LAN use requires an access token.</p>
+    <p>
+      Merge PDF or DOCX parts from Windows, macOS, Linux, Android, iPhone/iPad, ChromeOS,
+      or any modern browser.
+    </p>
+    <p class="muted">
+      Files are processed by the DocMergeForge Python host you connected to. The default server
+      binds only to this computer; LAN use requires an access token.
+    </p>
 
     <form id="merge-form">
       <label for="files">PDF or DOCX files</label>
-      <input id="files" name="files" type="file" accept=".pdf,.docx,application/pdf,application/vnd.openxmlformats-officedocument.wordprocessingml.document" multiple required>
+      <input
+        id="files"
+        name="files"
+        type="file"
+        accept=".pdf,.docx,application/pdf,application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+        multiple
+        required
+      >
 
       <div class="grid">
         <div>
@@ -83,8 +109,16 @@
       </div>
 
       <label for="access-token">Access token (LAN only)</label>
-      <input id="access-token" type="password" autocomplete="off" aria-describedby="access-token-help">
-      <p id="access-token-help" class="muted">Leave blank for the loopback-only host. LAN tokens stay in this browser tab session and are sent only in the merge request header.</p>
+      <input
+        id="access-token"
+        type="password"
+        autocomplete="off"
+        aria-describedby="access-token-help"
+      >
+      <p id="access-token-help" class="muted">
+        Leave blank for the loopback-only host. LAN tokens stay in this browser tab session and
+        are sent only in the merge request header.
+      </p>
 
       <button id="merge-button" type="submit">Merge and download</button>
       <div id="status" class="status" role="status" aria-live="polite"></div>
@@ -100,7 +134,9 @@
 </main>
 <script>
 (() => {
-  const fragment = new URLSearchParams(location.hash.startsWith("#") ? location.hash.slice(1) : location.hash);
+  const fragment = new URLSearchParams(
+    location.hash.startsWith("#") ? location.hash.slice(1) : location.hash
+  );
   const fragmentToken = fragment.get("token");
   if (fragmentToken) {
     sessionStorage.setItem("docmergeforge-token", fragmentToken);
@@ -176,7 +212,9 @@
 
 _SW = """const CACHE='docmergeforge-shell-v1';
 const SHELL=['/','/manifest.webmanifest','/icon.svg'];
-self.addEventListener('install', event => event.waitUntil(caches.open(CACHE).then(cache => cache.addAll(SHELL))));
+self.addEventListener('install', event =>
+  event.waitUntil(caches.open(CACHE).then(cache => cache.addAll(SHELL)))
+);
 self.addEventListener('activate', event => event.waitUntil(self.clients.claim()));
 self.addEventListener('fetch', event => {
   if (event.request.method !== 'GET') return;
@@ -200,9 +238,7 @@ def ordered_documents(input_root: Path) -> tuple[DocumentKind, list[InputDocumen
     """Discover one homogeneous PDF/DOCX upload set and order numbered parts naturally."""
 
     documents = [
-        item
-        for item in scan([input_root])
-        if item.kind in {DocumentKind.PDF, DocumentKind.DOCX}
+        item for item in scan([input_root]) if item.kind in {DocumentKind.PDF, DocumentKind.DOCX}
     ]
     if not documents:
         raise ValueError("No PDF or DOCX files were uploaded.")
@@ -249,7 +285,10 @@ async def _save_uploads(
             if suffix not in {".pdf", ".docx"}:
                 raise HTTPException(
                     status_code=415,
-                    detail=f"Unsupported upload type for {filename}. Only PDF and DOCX are accepted.",
+                    detail=(
+                        f"Unsupported upload type for {filename}. "
+                        "Only PDF and DOCX are accepted."
+                    ),
                 )
             destination_dir = input_root / f"{index:04d}"
             destination_dir.mkdir(parents=True, exist_ok=True)
@@ -425,4 +464,4 @@ async def merge(
                 detail="Merge failed. Check the DocMergeForge host logs for details.",
             ) from exc
 
-    return app
\ No newline at end of file
+    return app
diff --git a/tests/integration/test_lo_uno_process_group.py b/tests/integration/test_lo_uno_process_group.py
index 2e0dd43..a79e2a1 100644
--- a/tests/integration/test_lo_uno_process_group.py
+++ b/tests/integration/test_lo_uno_process_group.py
@@ -7,7 +7,6 @@
 
 from docmergeforge.docx import libreoffice_uno_merge
 
-
 pytestmark = pytest.mark.skipif(
     os.name != "posix",
     reason="LibreOffice UNO process-group acceptance currently uses POSIX semantics.",
diff --git a/tests/integration/test_lo_uno_supervised_smoke.py b/tests/integration/test_lo_uno_supervised_smoke.py
index af5bc39..b255d9c 100644
--- a/tests/integration/test_lo_uno_supervised_smoke.py
+++ b/tests/integration/test_lo_uno_supervised_smoke.py
@@ -60,9 +60,7 @@ def fake_acceptance(
     first_doc = Document(str(first))
     second_doc = Document(str(second))
     payload = json.loads(
-        (output_dir / "libreoffice-uno-merge-evidence.json").read_text(
-            encoding="utf-8"
-        )
+        (output_dir / "libreoffice-uno-merge-evidence.json").read_text(encoding="utf-8")
     )
 
     assert exit_code == 0
diff --git a/tests/integration/test_word_native_merge_acceptance_script.py b/tests/integration/test_word_native_merge_acceptance_script.py
index 85e01fc..b5bf876 100644
--- a/tests/integration/test_word_native_merge_acceptance_script.py
+++ b/tests/integration/test_word_native_merge_acceptance_script.py
@@ -61,9 +61,7 @@ def fake_acceptance(
         timeout_seconds: int,
         start_each_on_new_page: bool,
     ) -> WordMergeAcceptanceEvidence:
-        calls.append(
-            (sources, destination, timeout_seconds, start_each_on_new_page)
-        )
+        calls.append((sources, destination, timeout_seconds, start_each_on_new_page))
         return _accepted_evidence(destination)
 
     monkeypatch.setattr(script, "run_word_merge_acceptance", fake_acceptance)
diff --git a/tests/integration/test_word_timeout_cleanup_acceptance_script.py b/tests/integration/test_word_timeout_cleanup_acceptance_script.py
index a163562..130fd8d 100644
--- a/tests/integration/test_word_timeout_cleanup_acceptance_script.py
+++ b/tests/integration/test_word_timeout_cleanup_acceptance_script.py
@@ -34,9 +34,7 @@ def test_word_timeout_cleanup_acceptance_records_timeout_and_cleanup(
         lambda mode: _capability(),
     )
 
-    def fake_run(
-        command: list[str], *, timeout_seconds: int
-    ) -> NativeCommandResult:
+    def fake_run(command: list[str], *, timeout_seconds: int) -> NativeCommandResult:
         nonlocal captured_script
         identity = Path(command[command.index("-ProcessIdentityFile") + 1])
         identity.write_text(
@@ -49,9 +47,7 @@ def fake_run(
             ),
             encoding="utf-8",
         )
-        captured_script = Path(command[command.index("-File") + 1]).read_text(
-            encoding="utf-8"
-        )
+        captured_script = Path(command[command.index("-File") + 1]).read_text(encoding="utf-8")
         raise ValidationError(
             f"Native DOCX fidelity command timed out after {timeout_seconds} seconds."
         )
@@ -78,9 +74,7 @@ def fake_run(
         ]
     )
     payload = json.loads(
-        (output_dir / "word-timeout-cleanup-evidence.json").read_text(
-            encoding="utf-8"
-        )
+        (output_dir / "word-timeout-cleanup-evidence.json").read_text(encoding="utf-8")
     )
 
     assert exit_code == 0
@@ -103,9 +97,7 @@ def test_word_timeout_cleanup_acceptance_rejects_non_timeout_failure(
         lambda mode: _capability(),
     )
 
-    def fail_before_timeout(
-        command: list[str], **kwargs: object
-    ) -> NativeCommandResult:
+    def fail_before_timeout(command: list[str], **kwargs: object) -> NativeCommandResult:
         raise ValidationError("Word COM automation failed before timeout.")
 
     monkeypatch.setattr(script, "run_native_command", fail_before_timeout)
diff --git a/tests/unit/test_atomic.py b/tests/unit/test_atomic.py
index d280585..7128a87 100644
--- a/tests/unit/test_atomic.py
+++ b/tests/unit/test_atomic.py
@@ -76,9 +76,10 @@ def failing_fsync(fd: int) -> None:
 
     monkeypatch.setattr(atomic.os, "fsync", failing_fsync)
 
-    with pytest.raises(OSError, match="simulated fsync failure"), atomic_output(
-        target, overwrite=True
-    ) as temp:
+    with (
+        pytest.raises(OSError, match="simulated fsync failure"),
+        atomic_output(target, overwrite=True) as temp,
+    ):
         temp.write_bytes(b"new")
 
     assert target.read_bytes() == b"published"
diff --git a/tests/unit/test_docx_fidelity_corpus.py b/tests/unit/test_docx_fidelity_corpus.py
index 38e230e..8ae6981 100644
--- a/tests/unit/test_docx_fidelity_corpus.py
+++ b/tests/unit/test_docx_fidelity_corpus.py
@@ -68,9 +68,7 @@ def test_run_fidelity_corpus_keeps_report_paths_relative(
     monkeypatch.setattr(
         fidelity_corpus,
         "run_fidelity_roundtrip_acceptance",
-        lambda source, destination, mode, **kwargs: _accepted_evidence(
-            source, destination, mode
-        ),
+        lambda source, destination, mode, **kwargs: _accepted_evidence(source, destination, mode),
     )
     report = fidelity_corpus.run_fidelity_corpus(corpus, output, "libreoffice")
     payload = report.to_dict()
diff --git a/tests/unit/test_docx_section_evidence.py b/tests/unit/test_docx_section_evidence.py
index 83ddbd5..8393216 100644
--- a/tests/unit/test_docx_section_evidence.py
+++ b/tests/unit/test_docx_section_evidence.py
@@ -136,9 +136,9 @@ def test_page_number_fingerprint_binds_document_order(tmp_path: Path) -> None:
         chapter_separator="",
     )
 
-    assert page_number_properties_sha256(
-        [first, second]
-    ) != page_number_properties_sha256([second, first])
+    assert page_number_properties_sha256([first, second]) != page_number_properties_sha256(
+        [second, first]
+    )
 
 
 def test_page_number_evidence_rejects_non_docx(tmp_path: Path) -> None:
diff --git a/tests/unit/test_docx_word_merge.py b/tests/unit/test_docx_word_merge.py
index 34111c5..d9a7998 100644
--- a/tests/unit/test_docx_word_merge.py
+++ b/tests/unit/test_docx_word_merge.py
@@ -59,9 +59,7 @@ def fake_run(command: list[str], *, timeout_seconds: int) -> NativeCommandResult
         destination = Path(command[command.index("-Destination") + 1])
         sources = json.loads(manifest.read_text(encoding="utf-8"))
         captured["sources"] = sources
-        captured["script"] = Path(command[command.index("-File") + 1]).read_text(
-            encoding="utf-8"
-        )
+        captured["script"] = Path(command[command.index("-File") + 1]).read_text(encoding="utf-8")
         _write_identity_from_command(command)
 
         merged = Document()
@@ -74,9 +72,7 @@ def fake_run(command: list[str], *, timeout_seconds: int) -> NativeCommandResult
 
     cleanup_calls: list[tuple[Path, str]] = []
 
-    def fake_cleanup(
-        identity_file: Path, *, powershell: str
-    ) -> WordProcessCleanupResult:
+    def fake_cleanup(identity_file: Path, *, powershell: str) -> WordProcessCleanupResult:
         cleanup_calls.append((identity_file, powershell))
         return _clean_process_result()
 
@@ -157,9 +153,7 @@ def failing_run(command: list[str], *, timeout_seconds: int) -> NativeCommandRes
         _write_identity_from_command(command)
         raise TimeoutError("simulated PowerShell timeout")
 
-    def fake_cleanup(
-        identity_file: Path, *, powershell: str
-    ) -> WordProcessCleanupResult:
+    def fake_cleanup(identity_file: Path, *, powershell: str) -> WordProcessCleanupResult:
         cleanup_calls.append(identity_file)
         return WordProcessCleanupResult(
             identity_present=True,
diff --git a/tests/unit/test_docx_word_merge_acceptance.py b/tests/unit/test_docx_word_merge_acceptance.py
index a05072a..8e9bf3f 100644
--- a/tests/unit/test_docx_word_merge_acceptance.py
+++ b/tests/unit/test_docx_word_merge_acceptance.py
@@ -67,9 +67,7 @@ def _synthetic_merge(sources: tuple[Path, ...], output: Path) -> None:
     for index, source in enumerate(sources):
         current = Document(str(source))
         target_section = (
-            merged.sections[0]
-            if index == 0
-            else merged.add_section(WD_SECTION.NEW_PAGE)
+            merged.sections[0] if index == 0 else merged.add_section(WD_SECTION.NEW_PAGE)
         )
         _copy_section_layout(current.sections[0], target_section)
         for paragraph in current.paragraphs:
diff --git a/tests/unit/test_docx_word_merge_cleanup_failure.py b/tests/unit/test_docx_word_merge_cleanup_failure.py
index 1543d07..9766fd2 100644
--- a/tests/unit/test_docx_word_merge_cleanup_failure.py
+++ b/tests/unit/test_docx_word_merge_cleanup_failure.py
@@ -23,9 +23,7 @@ def test_word_native_merge_surfaces_cleanup_failure_over_original_timeout(
     output = tmp_path / "merged.docx"
     _write_docx(source)
 
-    def failing_run(
-        command: list[str], *, timeout_seconds: int
-    ) -> NativeCommandResult:
+    def failing_run(command: list[str], *, timeout_seconds: int) -> NativeCommandResult:
         identity = Path(command[command.index("-ProcessIdentityFile") + 1])
         identity.write_text(
             json.dumps(
diff --git a/tests/unit/test_docx_word_merge_page_number_acceptance.py b/tests/unit/test_docx_word_merge_page_number_acceptance.py
index 76b5594..20d59b7 100644
--- a/tests/unit/test_docx_word_merge_page_number_acceptance.py
+++ b/tests/unit/test_docx_word_merge_page_number_acceptance.py
@@ -28,9 +28,7 @@ def _inject_page_number_properties(path: Path, *, start: str, fmt: str) -> None:
     if marker not in document_xml:
         raise AssertionError("Expected section columns marker in generated DOCX fixture")
     page_number = f'<w:pgNumType w:start="{start}" w:fmt="{fmt}"/>'.encode()
-    members["word/document.xml"] = document_xml.replace(
-        marker, page_number + marker, 1
-    )
+    members["word/document.xml"] = document_xml.replace(marker, page_number + marker, 1)
     with ZipFile(path, "w", compression=ZIP_DEFLATED) as output:
         for name, payload in members.items():
             output.writestr(name, payload)
@@ -65,17 +63,13 @@ def _copy_section_layout(source: object, target: object) -> None:
         target_story.is_linked_to_previous = source_story.is_linked_to_previous
 
 
-def _merge_without_page_number_properties(
-    sources: tuple[Path, ...], output: Path
-) -> None:
+def _merge_without_page_number_properties(sources: tuple[Path, ...], output: Path) -> None:
     merged = Document()
     merged._body.clear_content()
     for index, source in enumerate(sources):
         current = Document(str(source))
         target_section = (
-            merged.sections[0]
-            if index == 0
-            else merged.add_section(WD_SECTION.NEW_PAGE)
+            merged.sections[0] if index == 0 else merged.add_section(WD_SECTION.NEW_PAGE)
         )
         _copy_section_layout(current.sections[0], target_section)
         for paragraph in current.paragraphs:
@@ -137,10 +131,7 @@ def fake_merge(
         evidence.expected_content.body_paragraphs_sha256
         == evidence.output_content.body_paragraphs_sha256
     )
-    assert (
-        evidence.expected_content.tables_sha256
-        == evidence.output_content.tables_sha256
-    )
+    assert evidence.expected_content.tables_sha256 == evidence.output_content.tables_sha256
     assert (
         evidence.expected_content.section_properties_sha256
         == evidence.output_content.section_properties_sha256
diff --git a/tests/unit/test_lo_uno_acceptance_workflow.py b/tests/unit/test_lo_uno_acceptance_workflow.py
index b12e9a2..40b38e4 100644
--- a/tests/unit/test_lo_uno_acceptance_workflow.py
+++ b/tests/unit/test_lo_uno_acceptance_workflow.py
@@ -3,12 +3,7 @@
 
 def _workflow_text() -> str:
     repository_root = Path(__file__).resolve().parents[2]
-    workflow = (
-        repository_root
-        / ".github"
-        / "workflows"
-        / "libreoffice-uno-acceptance.yml"
-    )
+    workflow = repository_root / ".github" / "workflows" / "libreoffice-uno-acceptance.yml"
     return workflow.read_text(encoding="utf-8")
 
 
diff --git a/tests/unit/test_ooxml_risk.py b/tests/unit/test_ooxml_risk.py
index ed54fc3..49f5e07 100644
--- a/tests/unit/test_ooxml_risk.py
+++ b/tests/unit/test_ooxml_risk.py
@@ -3,7 +3,6 @@
 
 from docmergeforge.validation.ooxml import risky_docx_constructs
 
-
 _CONTENT_TYPES = """<?xml version="1.0" encoding="UTF-8"?>
 <Types xmlns="http://schemas.openxmlformats.org/package/2006/content-types">
   <Default Extension="rels" ContentType="application/vnd.openxmlformats-package.relationships+xml"/>
diff --git a/tests/unit/test_output_transaction_hardening.py b/tests/unit/test_output_transaction_hardening.py
index f358618..1036c12 100644
--- a/tests/unit/test_output_transaction_hardening.py
+++ b/tests/unit/test_output_transaction_hardening.py
@@ -33,11 +33,9 @@ def _entry(
         "had_existing": had_existing,
         "backup_name": backup_name,
         "staged_size": len(staged_data) if staged_size is None else staged_size,
-        "staged_sha256": (
-            hashlib.sha256(staged_data).hexdigest()
-            if staged_sha256 is None
-            else staged_sha256
-        ),
+        "staged_sha256": hashlib.sha256(staged_data).hexdigest()
+        if staged_sha256 is None
+        else staged_sha256,
     }
 
 
diff --git a/tests/unit/test_scanner_exclusions.py b/tests/unit/test_scanner_exclusions.py
index 8cf186c..083607a 100644
--- a/tests/unit/test_scanner_exclusions.py
+++ b/tests/unit/test_scanner_exclusions.py
@@ -66,6 +66,11 @@ def test_recursive_iter_files_prunes_excluded_directory_before_descent(
     source.mkdir()
     excluded = source / "Master"
     content = source / "Content"
+    content.mkdir()
+    root_file = source / "root.txt"
+    content_file = content / "Part 1.docx"
+    root_file.write_text("root", encoding="utf-8")
+    content_file.write_text("content", encoding="utf-8")
     directory_names = ["Master", "Content"]
     walk_calls: list[tuple[Path, bool]] = []
 
@@ -75,15 +80,15 @@ def fake_walk(
         followlinks: bool,
     ) -> Iterator[tuple[str, list[str], list[str]]]:
         walk_calls.append((root, followlinks))
-        yield str(source), directory_names, ["root.txt"]
+        yield str(source), directory_names, [root_file.name]
         assert directory_names == ["Content"]
-        yield str(content), [], ["Part 1.docx"]
+        yield str(content), [], [content_file.name]
 
     monkeypatch.setattr(scanner.os, "walk", fake_walk)
 
     discovered = list(scanner.iter_files([source], exclude_roots=[excluded]))
 
-    assert discovered == [source / "root.txt", content / "Part 1.docx"]
+    assert discovered == [root_file, content_file]
     assert walk_calls == [(source, False)]
 
 
@@ -107,9 +112,7 @@ def test_non_recursive_iter_files_still_honors_excluded_root(tmp_path: Path) ->
     included.write_text("one", encoding="utf-8")
     excluded_file.write_text("two", encoding="utf-8")
 
-    discovered = list(
-        scanner.iter_files([source], recursive=False, exclude_roots=[excluded])
-    )
+    discovered = list(scanner.iter_files([source], recursive=False, exclude_roots=[excluded]))
 
     assert discovered == [included]
 
diff --git a/tests/unit/test_ui_resources.py b/tests/unit/test_ui_resources.py
index cc407c5..cc7906d 100644
--- a/tests/unit/test_ui_resources.py
+++ b/tests/unit/test_ui_resources.py
@@ -10,7 +10,7 @@
 
 def test_desktop_support_links_match_canonical_project_values() -> None:
     assert REPOSITORY_URL == "https://github.com/sanskarIN/DocMergeForge"
-    assert DOCS_URL == f"{REPOSITORY_URL}/tree/main/docs"
+    assert f"{REPOSITORY_URL}/tree/main/docs" == DOCS_URL
     assert BMC_URL == "https://buymeacoffee.com/sanskarIN"
     assert X_URL == "https://x.com/x_sanskarIN"
     assert BUSINESS_EMAIL == "sanskarin@outlook.in"
diff --git a/tests/unit/test_version_metadata.py b/tests/unit/test_version_metadata.py
index f52d97f..29b7c00 100644
--- a/tests/unit/test_version_metadata.py
+++ b/tests/unit/test_version_metadata.py
@@ -3,6 +3,8 @@
 
 from docmergeforge import __version__
 
+EXPECTED_RELEASE_VERSION = "2.8.5"
+
 
 def _pyproject() -> dict[str, object]:
     root = Path(__file__).resolve().parents[2]
@@ -16,6 +18,14 @@ def test_pyproject_version_matches_package_version() -> None:
     assert project["version"] == __version__
 
 
+def test_release_version_is_pinned_for_candidate() -> None:
+    pyproject = _pyproject()
+    project = pyproject["project"]
+    assert isinstance(project, dict)
+    assert project["version"] == EXPECTED_RELEASE_VERSION
+    assert __version__ == EXPECTED_RELEASE_VERSION
+
+
 def test_public_console_scripts_use_maintained_entry_points() -> None:
     pyproject = _pyproject()
     project = pyproject["project"]
diff --git a/what_changed.md b/what_changed.md
index 23fb69d..0001324 100644
--- a/what_changed.md
+++ b/what_changed.md
@@ -3,12 +3,132 @@
 This file records the **current** DocMergeForge development continuation. Earlier detailed records are preserved so this top-level checkpoint remains fast to review:
 
 - [`docs/history/what_changed-through-2026-08-18.md`](docs/history/what_changed-through-2026-08-18.md) — earlier repository development history;
-- [`docs/history/what_changed-through-2026-08-20-cross-platform.md`](docs/history/what_changed-through-2026-08-20-cross-platform.md) — the complete top-level record through the responsive cross-platform web/security continuation immediately preceding this one.
+- [`docs/history/what_changed-through-2026-08-20-cross-platform.md`](docs/history/what_changed-through-2026-08-20-cross-platform.md) — the complete top-level record through the responsive cross-platform web/security continuation immediately preceding desktop synchronization.
 
 A source change, test file, workflow definition, or commit is implementation/configuration evidence only. It is not represented as a passing CI, packaged-app, browser/device, accessibility, external-office, signing, notarization, or production-release result unless that exact evidence was observed.
 
+## 2026-08-24 — Version 2.8.5 release preparation
+
+### Goal
+
+Prepare DocMergeForge package/runtime/release metadata for `2.8.5`, remove stale continuation guidance, and close the remaining GitHub Actions generation mismatch without claiming unobserved release acceptance.
+
+This work is based on `main` checkpoint `aac5bf6d275991e21b68e15f5ad31f084fbc72e2` and is being developed on `release/2.8.5-prep` for review before merge.
+
+### Version metadata synchronized
+
+- `pyproject.toml` now declares `version = "2.8.5"`.
+- `src/docmergeforge/__init__.py` now exposes `__version__ = "2.8.5"`.
+- `tests/unit/test_version_metadata.py` retains package/runtime equality checking and now pins `EXPECTED_RELEASE_VERSION = "2.8.5"` so candidate drift fails explicitly.
+- The existing pre-alpha classifier remains conservative. The numeric version is not treated as proof of production readiness.
+
+### Project Sync Safety workflow aligned
+
+`.github/workflows/project-sync-safety.yml` was the remaining focused workflow using older GitHub Action generations. It now uses:
+
+```text
+actions/checkout@v7
+actions/setup-python@v7
+```
+
+This matches the maintained Node-24-era action generation documented elsewhere in the repository. Dependabot PR #3 and PR #4 propose the same two one-line upgrades independently; after this combined release-preparation change is merged, those PRs can be closed as superseded rather than merged separately.
+
+### Recent-project synchronization is complete
+
+The earlier continuation record listed a Recent Projects synchronization shortcut as possible future work. That item is now complete on `main`.
+
+`ProjectSyncMainWindow` exposes two stable accessible actions:
+
+- **Synchronize Project Sources** for browsing to a project JSON;
+- **Synchronize Recent Project** for selecting from maintained recent-project history.
+
+Both routes call the same `_synchronize_project_path(...)` workflow. They therefore preserve the shared `project.sync` planner/apply semantics, full preview, same-kind duplicate blocking, unchanged no-op behavior, separate removal approval, exact revision propagation, backup behavior, stale-write failure surfacing, and metadata-only mutation boundary.
+
+The integration suite covers both browse and recent-project routing in addition to the synchronization safety contract.
+
+### Project checkpoint synchronized
+
+`PROJECT_STATE.md` now records:
+
+- target `main` plus active `release/2.8.5-prep` branch;
+- package version `2.8.5`;
+- release-preparation base `aac5bf6d275991e21b68e15f5ad31f084fbc72e2`;
+- the Recent Projects shortcut as completed rather than future work;
+- exact candidate verification gates that still require observed evidence.
+
+### CI-discovered maintenance fixed without weakening gates
+
+The first candidate runs exposed pre-existing maintenance drift that was not visible from source inspection alone. The fixes stay inside the maintained contracts rather than suppressing checks:
+
+1. Project Sync Safety initially failed Black 26 formatting on four existing paths. The branch now applies Black's formatting-only result to:
+   - `src/docmergeforge/discovery/scanner.py`;
+   - `src/docmergeforge/project/sync.py`;
+   - `src/docmergeforge/cli/main.py`;
+   - `tests/unit/test_scanner_exclusions.py`.
+   A subsequent Ubuntu focused-safety run observed Black, Ruff, and strict mypy passing on the corrected files.
+2. The focused pruning regression then exposed an invalid fixture assumption: its fake `os.walk` yielded names for files/directories that had never been created, while production scanning correctly requires `Path.is_file()`. The test now creates the real root/content files before exercising pruning; production scanner behavior was not weakened.
+3. 120-Part Regression exposed two workflow collection defects:
+   - the integration suite includes FastAPI-backed web tests but the workflow installed only `[dev]`;
+   - several integration tests import repository-root `scripts`, but the workflow invoked the standalone `pytest` executable.
+   The workflow now installs `.[dev,web]` and invokes `python -m pytest`, preserving repository-root imports and the actual integration dependency set.
+4. Quality now likewise uses `python -m pip` and `python -m pytest` so its execution environment matches the repository import model deterministically.
+
+These are CI/test-environment corrections, not reductions in coverage or quality rules.
+
+### Focused commits in this continuation
+
+- `253aed48f661b5009c95d2a298cb7038081ac1f8` — `ci(sync): align focused workflow action majors`.
+- `4eda83909a62fb2fd6dbf6a3ca05da6e50b681a8` — `chore(version): prepare package metadata for 2.8.5`.
+- `1823ff8bf406055298dbdaed82f84a83d836196a` — `chore(version): expose 2.8.5 package version`.
+- `67015d621ca4b86ab4731b79dded7e0badd967e8` — `test(version): pin 2.8.5 release candidate metadata`.
+- `1ee0e76d6714cd4c2c15c02dc6c8edaeab918454` — `docs(state): advance checkpoint to 2.8.5 preparation`.
+- `a456bf41813f20110500a1d3b5b34773239c40c9` — `docs(progress): record 2.8.5 preparation work`.
+- `55039559d7f444ddfc528c1986f4de7deeea18d6` — `docs(release): define evidence-first 2.8.5 candidate policy`.
+- `3c7ca61672be48255cb159aec81fe98e7eb11f46` — `style(discovery): apply maintained Black formatting`.
+- `2eb2fbf448a91451169c37b037e317a4a7d1b323` — `style(sync): apply maintained Black formatting`.
+- `2d85c31ea341be976a7cacf341bf957b3ec7e9ce` — `style(test): format scanner exclusion regression`.
+- `c9f9c3995558d0f087ea692fa322e70ce5cdcf2b` — `style(cli): format validation diagnostics`.
+- `01d10892af0a5f3ba66f57fb32894587dcba8e8d` — `test(discovery): create real files for pruning regression`.
+- `1b03b5aa34092b8ee7dddf6480903502bd2cade7` — `ci(regression): install integration extras and preserve repo imports`.
+- `ae302a4ee7388d4f84b7c598e97aac7d4f9b0137` — `ci(quality): preserve repository imports during pytest`.
+
+### Verification boundary for 2.8.5
+
+Fresh candidate runs are required after the latest documentation checkpoint. Intermediate observed evidence is intentionally recorded only at the granularity actually seen:
+
+- Project Sync Safety on Ubuntu, after the formatting commits and before the later test/workflow commits, passed focused Black, Ruff, and strict mypy and then exposed the pruning-fixture failure described above.
+- Security dependency review was observed passing on a later candidate head while CodeQL was still running.
+- Earlier failed/interrupted runs are diagnostic evidence only and are superseded by newer candidate heads.
+
+Before `2.8.5` is treated as release-verified, observe and review the exact final candidate commit for the applicable gates, including:
+
+- Quality on Python 3.12 and 3.13;
+- Project Sync Safety on Ubuntu, Windows, and macOS;
+- 120-Part Regression;
+- Build Smoke on Ubuntu, Windows, and macOS;
+- Security/CodeQL;
+- Package Desktop and Onefile Acceptance if those artifacts are distributed;
+- downloaded-artifact verification;
+- release documentation/reference checks.
+
+Human/production gates remain independent: representative device/browser acceptance, clean-machine desktop UX, human accessibility, real-world office fidelity, claimed-scale stress evidence, Windows signing, macOS signing/notarization, and any installer/distribution acceptance.
+
+Historical workflow runs remain evidence only for their exact historical checkpoints; they are not relabeled as `2.8.5` evidence.
+
+### Remaining next work after this preparation pass
+
+1. Observe the exact final PR-head Quality, Project Sync Safety, 120-Part Regression, Build Smoke, and Security/CodeQL results and fix any remaining failure without weakening maintained gates.
+2. Finish optional public release-facing wording in `CHANGELOG.md`/README only where it adds accurate value; `PROJECT_STATE.md`, `what_changed.md`, package/runtime metadata, and release policy are already synchronized to the candidate boundary.
+3. Run/review Package Desktop, Onefile Acceptance, and downloaded-artifact gates if those artifacts are intended for the `2.8.5` distribution.
+4. Continue representative browser/device acceptance; browser support is not native APK/AAB/IPA packaging.
+5. Continue native-office fidelity, measured large-stress, human accessibility, clean-machine, signing/notarization, and distribution acceptance independently.
+6. Keep synchronized project discovery/business rules centralized in `project.sync`; do not fork browse/recent/CLI semantics.
+7. If simultaneous multi-writer project editing becomes required, design a coordinated lock/revision protocol rather than relabeling the optimistic SHA-256 stale-write guard.
+
 ## 2026-08-20 — Guarded desktop project synchronization
 
+> Historical continuation below. Its original “remaining next work” list is superseded by the `2026-08-24` section above where later commits completed the Recent Projects shortcut and began `2.8.5` preparation.
+
 ### Goal completed
 
 The desktop application now exposes the CLI project's review-first synchronization model instead of requiring desktop users to switch to the command line when a saved project's source membership changes.
@@ -50,24 +170,17 @@ The implementation deliberately **reuses** the existing `docmergeforge.project.s
 
 ### Changed
 
-- `pyproject.toml` now routes the public desktop console script through the maintained synchronization-enabled entry:
+- `pyproject.toml` routes the public desktop console script through the maintained synchronization-enabled entry:
 
   ```text
   docmergeforge-gui = "docmergeforge.ui.desktop_entry:main"
   ```
 
-- `src/docmergeforge/ui/packaged_entry.py` now:
-  - delegates normal packaged startup to the synchronization-enabled desktop entry;
-  - instantiates `ProjectSyncMainWindow` during packaged GUI smoke;
-  - retains the existing real temporary PDF/DOCX publication smoke after GUI initialization.
-- `src/docmergeforge/packaging/desktop.py` now requires all maintained desktop startup modules during `build_desktop.py --check`:
-  - `ui/main.py`;
-  - `ui/desktop_entry.py`;
-  - `ui/project_sync_dialog.py`;
-  - `ui/packaged_entry.py`.
-- `tests/unit/test_build_desktop.py` now protects those packaging prerequisites.
-- `tests/unit/test_version_metadata.py` now pins all maintained public console entry points so metadata drift cannot silently bypass the desktop synchronization entry.
-- The root `README.md` now publicly describes guarded desktop synchronization, its metadata-only boundary, duplicate blocking, second removal approval, backup behavior, and exact-revision stale-write protection.
+- `src/docmergeforge/ui/packaged_entry.py` delegates normal packaged startup to the synchronization-enabled desktop entry, instantiates `ProjectSyncMainWindow` during packaged GUI smoke, and retains the existing real temporary PDF/DOCX publication smoke after GUI initialization.
+- `src/docmergeforge/packaging/desktop.py` requires all maintained desktop startup modules during `build_desktop.py --check`: `ui/main.py`, `ui/desktop_entry.py`, `ui/project_sync_dialog.py`, and `ui/packaged_entry.py`.
+- `tests/unit/test_build_desktop.py` protects those packaging prerequisites.
+- `tests/unit/test_version_metadata.py` pins all maintained public console entry points so metadata drift cannot silently bypass the desktop synchronization entry.
+- The root `README.md` describes guarded desktop synchronization, its metadata-only boundary, duplicate blocking, second removal approval, backup behavior, and exact-revision stale-write protection.
 
 ### Shared synchronization safety preserved
 
@@ -100,11 +213,11 @@ Updated:
 - `docs/test-suite-reference.md`;
 - `docs/repository-reference-cross-platform.md`.
 
-The repository-reference addendum explicitly catalogs the new desktop runtime/test paths and this archived prior progress record so the tracked-file documentation contract remains maintainable.
+The repository-reference addendum catalogs the new desktop runtime/test paths and archived prior progress record so the tracked-file documentation contract remains maintainable.
 
 ### Packaging and entry-point hardening
 
-The new desktop feature is not source-only:
+The desktop feature is not source-only:
 
 - editable/wheel installs launch it through `docmergeforge-gui`;
 - packaged application startup launches the same synchronization-enabled window;
@@ -112,7 +225,7 @@ The new desktop feature is not source-only:
 - packaging preflight refuses a repository missing the new desktop entry/dialog modules;
 - metadata regression tests pin CLI, GUI, and web console targets.
 
-### Commits in this continuation before this record reset
+### Commits in that continuation before the record reset
 
 - `f6bb8ffcf72500655006c392844963fcb6459d03` — `feat(ui): add project synchronization preview dialog`.
 - `1ec59c4923961b8a34e8a3badaabc993937fc2c8` — `docs(reference): catalog desktop project sync dialog`.
@@ -137,24 +250,16 @@ The new desktop feature is not source-only:
 - `8ac078d1f0ccc5aa014cea4a551713cb0ff12046` — `docs(readme): expose guarded desktop project synchronization`.
 - `ceb27871b64c500121d2fb89da1317dc21b32171` — `docs(reference): catalog archived cross-platform progress record`.
 
-One intermediate `pyproject.toml` write accidentally added Ruff's `S` rule family while changing the GUI entry point. It was immediately reverted in the next focused commit (`b8328a7...`). The repository's maintained Ruff rule set remains `E/F/I/B/UP/SIM/C4`; this continuation does not weaken or silently broaden lint policy.
+One intermediate `pyproject.toml` write accidentally added Ruff's `S` rule family while changing the GUI entry point. It was immediately reverted in the next focused commit (`b8328a7...`). The maintained Ruff rule set remains `E/F/I/B/UP/SIM/C4`; that continuation did not weaken or silently broaden lint policy.
 
-### Verification status
+### Historical verification status
 
 - Continuation base: `9775190f38e613e33f20aafc82b678a1ca3a233d`.
-- Latest implementation/documentation checkpoint immediately before this active-record reset: `ceb27871b64c500121d2fb89da1317dc21b32171`.
-- A compare from the continuation base through the pre-record checkpoint shows the expected GUI/project-sync/packaging/test/documentation surfaces and no unrelated document-engine rewrite.
-- The public repository remains version `0.1.0` / pre-stable.
-- Branch metadata inspected during this continuation reports `main` as not protected and required status checks disabled at the repository-rules layer. This is repository-administration state, not an application defect; enabling branch protection remains an administrative decision.
-- The execution environment could not obtain a raw GitHub archive/checkout, so no local current-head Ruff, Black, mypy, docs checker, repository-reference checker, or pytest run is claimed.
-- GitHub workflow definitions remain configured to run Quality on `main`, but configured automation is not treated as an observed passing current-head result.
-
-### Remaining next work
-
-1. Observe a current-head Quality run and fix any Ruff/Black/mypy/docs/reference/pytest failure without weakening the checks.
-2. Review Build Smoke/package results for the synchronization-enabled desktop entry and fix any platform-specific startup or PyInstaller issue found by real runners.
-3. Consider a recent-project shortcut into synchronization only if it can preserve the same explicit preview/removal/stale-write approvals; do not bypass the current safety gates for convenience.
-4. Keep the desktop and CLI synchronization semantics shared in `project.sync`; do not fork them into UI-specific business logic.
-5. Perform representative manual browser/device acceptance for the responsive cross-platform client.
-6. Continue independent release gates for native-office fidelity, measured multi-gigabyte stress, human accessibility, clean-machine packaged apps, Windows signing, and macOS signing/notarization.
-7. Do not claim `v1.0.0`, native mobile APK/AAB/IPA delivery, universal multi-writer project locking, or production certification until corresponding implementation and acceptance evidence exists.
+- Pre-record checkpoint: `ceb27871b64c500121d2fb89da1317dc21b32171`.
+- At that checkpoint the package metadata was still `0.1.0`; the later `2026-08-24` continuation above supersedes that version state.
+- Branch metadata reported `main` as not protected and required status checks disabled at the repository-rules layer.
+- No current-candidate test result should be inferred from the existence of committed tests/workflows.
+
+### Historical next-work list
+
+The original list included current-head Quality/build/package review, the then-unimplemented Recent Projects shortcut, manual browser/device acceptance, shared synchronization semantics, and independent release gates. The shortcut is now complete; the remaining applicable gates are carried forward in the `2026-08-24` section above.