YuzeJ21 · YuzeJ21 · Aug 10, 2026 · Aug 10, 2026 · Aug 10, 2026 · Aug 10, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -9,7 +9,28 @@ its linked release entry for the exact published source and assets.
 
 ## Unreleased
 
-No changes currently recorded.
+Development version: `0.2.4.dev0`. Public install remains the immutable v0.2.3 release.
+
+### Post-release engineering
+
+- Added CLI lifecycle parity for criterion resolution, atomic external E3/E4 runtime-evidence
+  recording, final acceptance, and changed-head comparison over validated local records.
+- Enforced a strict saved-record envelope and lifecycle-output validation. Failed commands do not
+  mutate saved records, and low-evidence acceptance notes share one fail-closed core policy across
+  CLI and Streamlit.
+- Added an installed-wheel packaged Chromium regression at 1280×720 and 390×844 with exact
+  Playwright 1.62.0, isolated local storage, loopback-only networking, and console/page-error
+  assertions.
+- Added Python 3.13 package, CLI, deterministic-benchmark, and exact workbench-health engineering
+  coverage in both a genuine local CPython 3.13 environment and protected CI.
+- Added a keyboard-only installed-workbench path with visible-focus assertions and
+  bounded native 200% zoom evidence on the tested macOS/Chrome configuration.
+- Enforced verified-public provenance before a live GitHub source can be persisted, exported, or
+  counted toward Alpha. Private, ambiguous, malformed, and legacy-unverified sources fail closed.
+
+These changes are ScopeProof engineering evidence only and earn zero Stage 1 credit. Real
+screen-reader operation, Windows desktop, Linux desktop, non-Chromium browser behavior, and WCAG
+conformance remain unsupported.
 
 ## 0.2.3 — Evidence integrity and reviewer loop
 

diff --git a/README.md b/README.md
@@ -111,6 +111,12 @@ at zero. The
 [exact-head verification audit](docs/audits/exact-head-runtime-evidence/verification.md)
 records the product-tree engineering evidence and remaining gaps.
 
+The repository's current source line is the unreleased development version `0.2.4.dev0`; it is
+not a replacement release asset. Post-v0.2.3 engineering merged CLI lifecycle commands, strict
+saved-record validation, installed Chromium coverage, Python 3.13 and bounded keyboard/focus
+evidence, and verified-public provenance enforcement. Those changes remain engineering evidence
+only. The public install continues to be the immutable v0.2.3 release, and Stage 1 remains zero.
+
 GitHub exposes visible check runs but does not reliably expose every repository's required-check
 policy to anonymous clients. ScopeProof therefore labels this value **Observed CI state** and counts
 only explicit success as passing.

diff --git a/ROADMAP.md b/ROADMAP.md
@@ -18,6 +18,7 @@ which provides the wheel, source archive, and checksum manifest.
 | Area | Current state |
 |---|---|
 | Published install | v0.2.3 GitHub Release with wheel, source archive, and checksum manifest |
+| Active source line | Unreleased `0.2.4.dev0`; no v0.2.4 release, tag, or package publication exists |
 | Verified product baseline | PR #184 release integration landed on `main` at `448c42758ea139bf9203cbf1bb04b02b02ae412c` |
 | Product verification | Full product-code verification is bound to `fb74d4bbb402f4de3e2fabb56ce28c948214f8c2`; package, install, installed-benchmark, and health artifacts are bound to `81598899fcd85df58ab22f9212f2e8382f4a5e5f`. |
 | Release integration evidence | PR #184 release integration at `448c42758ea139bf9203cbf1bb04b02b02ae412c`; exact-main CI, CodeQL, and Pages all succeeded, and `origin/main` matched at the 2026-08-08 branch-start snapshot |
@@ -28,6 +29,14 @@ which provides the wheel, source archive, and checksum manifest.
 Verify live GitHub and current release records before relying on publication state. Engineering
 milestones can proceed while Stage 1 waits, but they do not advance product-validation stages.
 
+After the v0.2.3 release, PR #185 merged core-backed CLI lifecycle parity and packaged Chromium
+proof at `30177733ef312ced22e6a2e57e3df6fdb1e92507`; PR #187 merged Python 3.13 plus bounded
+keyboard/focus and zoom engineering evidence at `c548759b5464ad5bb98baf1e996397f241dfc455`; and
+PR #188 merged verified-public provenance enforcement at
+`077f9351283b319b82854ad1df95eac7ce614e21`. CLI lifecycle parity is implemented;
+verified-public provenance enforcement is implemented. These changes form the `0.2.4.dev0`
+development line and earn no Stage 1 credit.
+
 ## Stage 0 — Reviewer-first product reset
 
 Status: prior PR #177 repairs remain historical engineering evidence. PR #180

diff --git a/docs/audits/accessibility-platform-evidence/verification.md b/docs/audits/accessibility-platform-evidence/verification.md
@@ -1,5 +1,9 @@
 # Accessibility and platform evidence verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 ## Evidence boundary
 
 - Date: 2026-08-09 (America/Toronto).

diff --git a/docs/audits/exact-head-runtime-evidence/verification.md b/docs/audits/exact-head-runtime-evidence/verification.md
@@ -1,5 +1,9 @@
 # Exact-head merged-product runtime-evidence verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 Machine-readable evidence manifest:
 [`verification.json`](verification.json). The manifest is the structured
 repository-contract input; this document remains the human-readable audit.

diff --git a/docs/audits/post-release-cli-browser/verification.md b/docs/audits/post-release-cli-browser/verification.md
@@ -1,5 +1,9 @@
 # Post-release truth, CLI parity, and packaged-browser verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 ## Evidence boundary
 
 - Date: 2026-08-08 (America/Toronto).

diff --git a/docs/audits/v0.2.3-integrity-reviewer-loop/verification.md b/docs/audits/v0.2.3-integrity-reviewer-loop/verification.md
@@ -1,5 +1,9 @@
 # ScopeProof v0.2.3 integrity and reviewer-loop verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 ## Evidence identity and boundary
 
 - Date: 2026-08-03 (America/Toronto)

diff --git a/docs/audits/v0.2.3-product-convergence/verification.md b/docs/audits/v0.2.3-product-convergence/verification.md
@@ -1,5 +1,9 @@
 # ScopeProof v0.2.3 product-convergence verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 ## Evidence identity and boundary
 
 - Date: 2026-08-02 (America/Toronto)

diff --git a/docs/audits/v0.2.3-workbench/accessibility-and-first-use-audit.md b/docs/audits/v0.2.3-workbench/accessibility-and-first-use-audit.md
@@ -1,5 +1,9 @@
 # v0.2.3 Workbench Accessibility and First-Use Audit
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 Status: internal engineering audit
 Date: 2026-07-26
 Environment: macOS, Streamlit local server, in-app Chromium browser

diff --git a/docs/audits/workbench-ux-simplification/verification.md b/docs/audits/workbench-ux-simplification/verification.md
@@ -1,5 +1,9 @@
 # Workbench UX simplification verification
 
+> **Historical evidence boundary:** This audit remains bound to its named commit, tree, version,
+> and environment. The [current status](../../releases/v0.2.3-status-and-next-stages.md) supersedes
+> unqualified present-state inferences and does not rewrite the results below.
+
 ## Scope and evidence boundary
 
 - Date: 2026-08-01 (America/Toronto)

diff --git a/docs/commercialization/market-comparison-2026-07-26.md b/docs/commercialization/market-comparison-2026-07-26.md
@@ -6,6 +6,9 @@ Boundary: competitor capabilities below are vendor-advertised unless explicitly
 labelled as a ScopeProof implementation fact. This document is not customer or
 market validation.
 
+Version boundary: public install remains v0.2.3; current source is the unreleased `0.2.4.dev0`
+development line. Post-release engineering changes do not constitute market validation.
+
 ## Product category
 
 ScopeProof is not a general AI code reviewer, test-management system, or static
@@ -87,6 +90,10 @@ criterion is supported at this exact head.”
 - Conservative False Ready posture and inspectable missing-evidence
   explanations.
 - Portable exports and versioned local records.
+- CLI lifecycle parity is implemented for core-backed resolution, external runtime verification,
+  final acceptance, and changed-head comparison.
+- Bounded keyboard-only and visible-focus engineering evidence is implemented for the installed
+  Chromium path; it is not accessibility conformance.
 
 ### Where ScopeProof is immature
 
@@ -97,8 +104,9 @@ criterion is supported at this exact head.”
   integration with an issue or test-management source.
 - Retrieval coverage is intentionally conservative and the completed R-002
   baseline found candidates for only 5 of 20 research criteria.
-- Accessibility is not yet verified with keyboard-only completion, a screen
-  reader, 200% zoom, Windows, or Linux.
+- Real screen-reader operation, Windows desktop, Linux desktop, non-Chromium browser behavior, and
+  WCAG conformance remain unsupported. The bounded keyboard/focus and native-zoom checks do not
+  establish those broader claims.
 - There is no evidence yet that users will repeat the workflow or pay for a team
   product.
 
@@ -112,7 +120,8 @@ criterion is supported at this exact head.”
 2. Preserve retrieval diagnostics as explanations, never verdict evidence.
 3. Improve conservative retrieval only through new constructed regressions and
    a prospectively frozen holdout; do not retune on R-002.
-4. Finish keyboard, zoom, assistive-technology, and available-platform checks.
+4. Attempt real assistive-technology and available desktop-platform checks only in genuine,
+   observable environments; keep unavailable rows unsupported.
 5. Keep export, changed-head re-review, and missing-evidence explanations clear.
 
 ### Design now, implement only after genuine use evidence

diff --git a/docs/development-environment.md b/docs/development-environment.md
@@ -1,6 +1,14 @@
 # Reproducible development environment
 
-ScopeProof supports Python 3.11 and newer. The contributor baseline is Python 3.12 with dependencies resolved by the checked-in `uv.lock`. This path uses only local and free open-source tooling; it does not require an OpenAI or other paid LLM API.
+The public install remains v0.2.3; the current repository source is the unreleased `0.2.4.dev0`
+development line. Python 3.11, Python 3.12, and Python 3.13 have current package/CLI engineering
+coverage. Python 3.11 is the declared floor, Python 3.12 is the locked contributor baseline, and
+Python 3.13 is exercised in protected compatibility CI and a genuine local interpreter check.
+Python 3.14 is unverified pending a clean, genuine compatibility run; the current `>=3.11`
+metadata must not be read as verified 3.14 support.
+
+Dependencies are resolved by the checked-in `uv.lock`. This path uses only local and free
+open-source tooling; it does not require an OpenAI or other paid LLM API.
 
 ## Create or refresh the environment
 
@@ -68,7 +76,12 @@ Run the local workbench with:
 uv run scopeproof-web --host 127.0.0.1 --port 8501
 ```
 
-The Python 3.11 CI lane remains the compatibility floor. A separate locked Python 3.12 lane verifies that the committed resolution can be recreated and runs repository contracts plus the deterministic benchmark before the required `verify` job.
+The Python 3.11 CI lane remains the compatibility floor. A separate locked Python 3.12 lane
+verifies that the committed resolution can be recreated and runs repository contracts plus both
+deterministic benchmarks. The Python 3.13 lane runs the complete suite, builds and installs a wheel,
+checks dependencies and both CLI versions, runs both installed benchmarks, and requires exact
+loopback workbench health before the required `verify` job. These Linux-runner checks are package
+and CLI evidence, not Linux desktop evidence.
 
 ## Known-good UI baseline
 

diff --git a/docs/releases/v0.2.3-platform-package-matrix.md b/docs/releases/v0.2.3-platform-package-matrix.md
@@ -5,12 +5,27 @@ Merged product commit: `2a320df966eff30c05a2b1dce607a247201fa165` (PR #180)
 Merged product tree: `add81a2d0ba7e64f8e4318a1959bbe7e6e4acfc8`
 Independently verified PR head: `ed9f9c0cf6b7cf7cc25403d6138e7a8391f55e0f`
 Verified PR-head tree: `add81a2d0ba7e64f8e4318a1959bbe7e6e4acfc8`
-Public install and latest release: v0.2.1
+Historical public install at the 2026-08-02 audit time: v0.2.1
+Current public install: v0.2.3
 Historical local checkout verification head: `87e95a76ca4100458465da520049992e8b39af5c`
 
 Historical package build base: `482d1a78f5d3345029d35d33a518ca36bc2f8e29`
 Classification: internal engineering evidence only
 
+## 2026-08-09 post-release development boundary
+
+The public package and every historical hash below remain bound to v0.2.3 and their named source
+trees. Current repository source is the unreleased `0.2.4.dev0` development line. Later merged
+engineering work added an installed-wheel Chromium lifecycle regression, a keyboard-only path with
+visible-focus assertions at 1280×720 and 390×844, bounded native 200% zoom on one macOS/Chrome
+configuration, and clean Python 3.13 package, CLI, benchmark, and loopback-health checks.
+
+Those results do not establish real screen-reader operation, Windows desktop, Linux desktop,
+non-Chromium browser behavior, or WCAG conformance. Hosted Linux CI is package/runtime evidence,
+not Linux desktop evidence. This additive boundary supersedes only unqualified current-status
+inferences; it does not rewrite any earlier measurement, artifact identity, or limitation below.
+All engineering evidence adds zero Stage 1 credit.
+
 ## 2026-08-02 merged product-tree package verification
 
 PR #180 merged the independently verified head's exact product tree as

diff --git a/docs/releases/v0.2.3-status-and-next-stages.md b/docs/releases/v0.2.3-status-and-next-stages.md
@@ -1,8 +1,9 @@
 # ScopeProof v0.2.3 status, gaps, and next stages
 
-Date: 2026-08-03
+Release-baseline date: 2026-08-03; current development alignment: 2026-08-09
 Current release baseline: PR #184 release integration at
 `448c42758ea139bf9203cbf1bb04b02b02ae412c`
+Current development line: unreleased `0.2.4.dev0` after merged PR #185, PR #187, and PR #188
 Exact-main engineering checks: CI run `30854382641`, CodeQL run `30854382413`,
 and Pages run `30854382659`; exact-main CI, CodeQL, and Pages all succeeded
 Historical source integration: PR #183 at
@@ -18,6 +19,15 @@ which contains the wheel, source archive, and checksum manifest.
 Product-validation state: Stage 0 source integration is complete; release and
 test work do not change Stage 1, which remains at zero. Stages 2–4 remain gated.
 
+PR #185 merged CLI lifecycle parity, the strict saved-record envelope, and packaged Chromium
+regression at `30177733ef312ced22e6a2e57e3df6fdb1e92507`. PR #187 merged Python 3.13 plus bounded
+keyboard/focus and native-zoom engineering evidence at
+`c548759b5464ad5bb98baf1e996397f241dfc455`. PR #188 merged verified-public provenance enforcement
+at `077f9351283b319b82854ad1df95eac7ce614e21`, so private, ambiguous, malformed, and
+legacy-unverified live sources fail closed. CLI lifecycle parity is implemented;
+verified-public provenance enforcement is implemented. These post-release changes belong to
+`0.2.4.dev0`, not the immutable v0.2.3 assets, and add zero Stage 1 credit.
+
 PR #174 head `e96d9929e651ca36bf178a6fedff0e89e8df3675` passed Python 3.11
 compatibility, locked-environment, `verify`, and CodeQL before it was merged to
 `main` as `178df8a1c8034edc4d3240d790e3e4c668dfa82a`. The merged source includes
@@ -120,7 +130,7 @@ code-review comments, scan security, or automatically fix a PR.
 | Exports | Pydantic-backed JSON, Markdown, CSV, and HTML with runtime identity, linked/unlinked state, and exact criteria-source provenance | Exported content is a review record, not certification |
 | Alpha evidence | One-time outcome capture only from a fully revalidated saved review with matching public-GitHub origin, PR, exact head, criteria, and provenance | Demo, fixture, research, legacy-unknown, mutated, or non-public origins contribute zero qualifying alpha evidence |
 | Engineering benchmarks | 12-case constructed acceptance benchmark, two-case comparison benchmark, R-001 research case, and frozen 20-case R-002 research baseline | All contribute zero Stage 1 credit |
-| Local workbench and CLI | Cleaner five-stage Streamlit hierarchy plus offline source-confirmation preparation, review, export, list, delete, benchmark, and comparison commands | Local owner operation and UI tests are not participant or accessibility validation |
+| Local workbench and CLI | Cleaner five-stage Streamlit hierarchy plus offline source-confirmation preparation; review, resolve, atomic runtime verification, final acceptance, export, list, delete, benchmark, and changed-head comparison commands | Local owner operation and UI tests are not participant or accessibility validation |
 
 ## Primary use cases
 
@@ -166,7 +176,7 @@ remains zero independent use.
 | 200% zoom | Supported bounded local evidence | Installed Chrome 151.0.7922.77 reported native `Zoom: 200%`, device pixel ratio changed from 1 to 2, the exact-head keyboard path completed, and no named control was horizontally clipped on the tested macOS/display configuration. This is not responsive resizing or a broad browser/platform claim. |
 | VoiceOver or another real screen reader | Unsupported current evidence | The installed VoiceOver service was detectable, but bounded control attempts exposed no observable screen-reader state, speech, caption, focus announcement, or reading-order result. |
 | Python 3.11 | Remote CI evidence only | Python 3.11 CI passed at the exact merged PR head; no local Python 3.11 desktop flow was executed |
-| Python 3.13 | Supported local engineering evidence | A genuine CPython 3.13.14 environment passed clean wheel installation, `pip check`, both versioned CLIs, both deterministic benchmarks, and exact workbench health. The new hosted 3.13 job must pass before its Linux-runner result is claimed. |
+| Python 3.13 | Supported package/runtime engineering evidence | A genuine local CPython 3.13.14 environment passed clean wheel installation, `pip check`, both versioned CLIs, both deterministic benchmarks, and exact workbench health. Protected Python 3.13 CI passed on PR #187 and resulting `main`; this is package/runtime evidence, not Linux desktop evidence. |
 | Windows and Linux desktop | Unavailable environment | Build, install, launch, and complete representative workflows on those operating systems |
 | Stage 1 | External evidence gate | A non-owner supplies a real public PR, public requirements, source-owner confirmation, exact head SHA, saved review, and genuine outcome |
 | Stage 2 | Stage gate | Every Stage 1 target passes before repeat-use and commercial-discovery evidence is collected |
@@ -250,20 +260,14 @@ target repository code.
    current PR head, invalidate stale prior conclusions on `synchronize`, and
    keep the current Action informational until independent use justifies
    promotion.
-2. **CLI lifecycle parity:** expose resolution, atomic external verification,
-   final acceptance, and changed-head comparison through the core-backed CLI
-   without weakening Streamlit or schema boundaries.
-3. **Non-executing evidence adapters:** design validated import records for
+2. **Non-executing evidence adapters:** design validated import records for
    JUnit-style results, coverage summaries, contract reports, build/deployment
    records, and externally supplied runtime attestations. Imported data remains
    evidence with provenance, not proof of correctness.
-4. **Authenticated reviewer identity:** current runtime records now retain
+3. **Authenticated reviewer identity:** current runtime records now retain
    review-scoped identity and attribution, but authentication, optional expiry,
    and externally signed attestations remain future decisions without accounts,
    silent overrides, or automatic approval.
-5. **Real-browser regression coverage:** automate the packaged pointer path
-   first, then add keyboard, responsive viewport, and zoom checks only where
-   the execution environment provides trustworthy control.
 
 ### Demand-gated candidates