KeyffMS · KeyffMS · Sep 8, 2026 · Sep 8, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,22 +4,30 @@ All notable changes to PlanAnvil are documented here.
 
 ## [Unreleased]
 
-### Qualification closure — 2026-09-07
-
-- full self-hosted run #27 (`34140846679`) reproduced all C01-C16 on executed source `a9cdcdc1e0cad70e88b60869e75e4166046dd306`, Codex CLI 0.153.4, `gpt-5.6-sol`, Debian 13;
-- finite C08 now proves the intended missing-checkpoint stop and completed repaired recovery without timeout; C09/C10 regressions passed;
-- preserve the exact full archive and provenance, update current capability evidence, and retain historical #25 unchanged;
-- C13 retains the documented project-scoped, known-error-gated non-ephemeral fallback; no new transport claim or product/runtime/security change;
-- qualification is recorded; signed production publication remains a separate action.
+## [0.2.0] - Pending publication
 
 ### Added
 
+- standard-library repository distribution manager with install, verify, upgrade, status and uninstall operations;
+- transactional rollback, ownership/hash state, conservative Codex config merging, and structural hook merging;
+- distribution tests covering clean repositories, existing `AGENTS.md`, existing `.codex/config.toml`, unrelated hooks, upgrade conflicts and uninstall conflicts;
+- deterministic release archive builder, candidate/production release gate, and tag-driven GitHub Release workflow;
+- deterministic C01-C16 qualification template archive containing fixture, prompt, config, expected result, current sanitized BLOCKED result, evaluation and SHA-256 manifests;
+- capability evidence materializer/validator/rehash tools and a live Codex sandbox runbook;
+- installation, troubleshooting, and release documentation;
 - add a controlled `workflow_dispatch`-only Codex qualification workflow for the trusted `plananvil-codex` self-hosted runner;
 - add a sequential C01-C16 live qualification controller that creates disposable fixture repositories, runs fresh ephemeral Codex trials, sanitizes structural evidence, rehashes/validates each package, and exports evidence only as a GitHub Actions artifact;
 - add regression tests ensuring the live qualification controller redacts private-looking data and never disables the Codex sandbox/approval boundary.
 
 ### Changed
 
+- refresh the Codex capability baseline through 2.2 (2026-08-28) to the qualified baseline 2.3 described below;
+- replace legacy agent concurrency/depth configuration with `agents.enabled` and `agents.max_concurrent_threads_per_session`;
+- enforce flat direct-child execution topology in the generated contract instead of relying on undocumented `agents.max_depth` behavior;
+- record the 2026-08-28 C01–C16 qualification attempt and its live Codex runtime blocker;
+- pin GitHub Actions to immutable SHAs and Node-24-based checkout/setup-python releases;
+- expand CI across Python 3.11 and the current upper supported interpreter on Ubuntu, macOS and Windows;
+- split distribution/release-candidate validation into a stable named CI check;
 - update pinned `actions/checkout` and `actions/setup-python` workflow SHAs to the current v7 releases while retaining immutable action pinning and Node 24 compatibility;
 - require the full Linux Codex qualification job to pass a system-`bubblewrap` user-namespace probe before C01-C16, so incompatible Podman runners fail fast instead of timing out capability-by-capability;
 - update the qualification evidence uploader to the Node-24-native `actions/upload-artifact` v6 immutable SHA;
@@ -33,38 +41,23 @@ All notable changes to PlanAnvil are documented here.
 - fail the production release gate closed when the release worktree is dirty or Git cleanliness cannot be verified;
 - document the controlled self-hosted Codex qualification path and keep the previous sandbox procedure as a manual fallback.
 
-## [0.2.0] - 2026-08-28
-
-### Added
-
-- standard-library repository distribution manager with install, verify, upgrade, status and uninstall operations;
-- transactional rollback, ownership/hash state, conservative Codex config merging, and structural hook merging;
-- distribution tests covering clean repositories, existing `AGENTS.md`, existing `.codex/config.toml`, unrelated hooks, upgrade conflicts and uninstall conflicts;
-- deterministic release archive builder, candidate/production release gate, and tag-driven GitHub Release workflow;
-- deterministic C01-C16 qualification template archive containing fixture, prompt, config, expected result, current sanitized BLOCKED result, evaluation and SHA-256 manifests;
-- capability evidence materializer/validator/rehash tools and a live Codex sandbox runbook;
-- installation, troubleshooting, and release documentation.
-
-### Changed
-
-- refresh the Codex capability baseline to 2.2 against current 2026-08-28 official documentation;
-- replace legacy agent concurrency/depth configuration with `agents.enabled` and `agents.max_concurrent_threads_per_session`;
-- enforce flat direct-child execution topology in the generated contract instead of relying on undocumented `agents.max_depth` behavior;
-- record the 2026-08-28 C01–C16 qualification attempt and its live Codex runtime blocker;
-- pin GitHub Actions to immutable SHAs and Node-24-based checkout/setup-python releases;
-- expand CI across Python 3.11 and the current upper supported interpreter on Ubuntu, macOS and Windows;
-- split distribution/release-candidate validation into a stable named CI check.
-
 ### Fixed
 
 - canonicalize event repository paths before active-run routing so source-worktree matching is stable across macOS symlink aliases and Windows path aliases;
 - keep checkpoint recovery assertions platform-neutral by comparing canonical paths;
 - update the artifact-sealing lock regression test to observe the current `validate_plan_contract` gate;
 - make the Git-hook probe fixture emit explicit hook diagnostics while preserving fail-closed classification for unrelated commit failures;
-- synchronize golden blind-review fixtures and dependent comparison hashes with the required independent `plan-anvil-reviewer` author role.
+- synchronize golden blind-review fixtures and dependent comparison hashes with the required independent `plan-anvil-reviewer` author role;
+- reconcile the compliance checklist and current qualification record with full run #27, retaining the historical #25 timeout and current C13 limitation;
+- check compliance metadata and finite C08 checklist status against the committed qualification index and summary in CI.
 
 ### Qualification closure — 2026-09-07
 
+- full self-hosted run #27 (`34140846679`) reproduced all C01-C16 on executed source `a9cdcdc1e0cad70e88b60869e75e4166046dd306`, Codex CLI 0.153.4, `gpt-5.6-sol`, Debian 13;
+- finite C08 now proves the intended missing-checkpoint stop and completed repaired recovery without timeout; C09/C10 regressions passed;
+- preserve the exact full archive and provenance, update current capability evidence, and retain historical #25 unchanged;
+- C13 retains the documented project-scoped, known-error-gated non-ephemeral fallback; no new transport claim or product/runtime/security change;
+- qualification is recorded; signed production publication remains a separate action;
 - preserve complete baseline 2.3 live evidence from full run #25, `34060321283`, tested at `d0384f76bc4150d33bb8f51ef5981f3243b3cfb3` with Codex CLI 0.153.4, `gpt-5.6-sol`, Debian 13; all C01–C16 were reproduced;
 - retain the exact source-bound archive, hashes and original limitations; C13 passed via the permitted project-native non-ephemeral fallback;
 - replace the old C08 repaired-path workload with a finite pressure/finish scenario and strict termination checks; the old positive timeout remains in historical evidence; full run #27 subsequently confirmed the replacement live;
@@ -74,7 +67,7 @@ All notable changes to PlanAnvil are documented here.
 
 ### Release status
 
-0.2.0 is qualified for the recorded configuration. Full run #27 (`34140846679`) confirmed C01-C16 and finite C08 stop/repair on source `a9cdcdc1e0cad70e88b60869e75e4166046dd306`; its complete source-bound evidence is committed and #25 remains unchanged. Production publication remains a separate verified signed annotated-tag action after strict release validation. No tag or release has been published by this evidence import.
+0.2.0 is qualified for the recorded configuration. Full run #27 (`34140846679`) confirmed C01-C16 and finite C08 stop/repair on source `a9cdcdc1e0cad70e88b60869e75e4166046dd306`; its complete source-bound evidence is committed and #25 remains unchanged. Production publication remains pending a verified signed annotated tag after strict release validation and green protected CI. No production tag or release is created by this documentation update.
 
 ## [0.1.0] - 2026-07-12
 

diff --git a/docs/OPENAI_COMPLIANCE.md b/docs/OPENAI_COMPLIANCE.md
@@ -1,6 +1,6 @@
 # PlanAnvil — OpenAI Codex Compliance Record
 
-> **Original contract review:** 2026-07-12; scoped recovery/source review: 2026-09-05; live qualification: 2026-09-06
+> **Original contract review:** 2026-07-12; scoped recovery/source review: 2026-09-05; live qualification: 2026-09-07; qualification-record reconciliation: 2026-09-08
 > **Scope:** contract 2.1; baseline 2.3 qualification of the 0.2.0 release candidate
 > **Rule:** current official OpenAI documentation is authoritative for Codex behavior.
 
@@ -199,13 +199,70 @@ A relevant documentation change blocks release until the affected record and tes
 - [x] Local profile and local state are ignored
 - [x] Machine-state formats defined
 - [x] Unsupported behaviors excluded from the active contract
-- [x] Complete source-scoped live C01–C16 evidence committed for run #25
-- [ ] Finite C08 repaired-path completion verified live and committed for production closure
-
-## 7. Source-scoped qualification and release closure — 2026-09-07
-
-Run `34060321283` reproduced all baseline 2.3 capabilities with Codex CLI `0.153.4`, model `gpt-5.6-sol`, Debian 13, at PlanAnvil `d0384f76bc4150d33bb8f51ef5981f3243b3cfb3`. The archived result retains its original source identifier and hashes. Current status and remaining constraints are in `QUALIFICATION_STATUS.md`; earlier audits are historical, not current blockers by default.
-
-C10 delivers recovery through `SessionStart(source=compact)`; `PostCompact` remains advisory. C13's project-native agent and hook pass via the allowed non-ephemeral retry, not an unrestricted home-scoped replacement. C08's historical positive trial timed out after proving unblocking. The stricter finite C08 follow-up has not yet produced live evidence and remains a production-publication prerequisite.
-
-The C08 fixture now isolates the deliberate invalid-checkpoint trigger from ordinary startup recovery by narrowing SessionStart to `^compact$` in the disposable root checkout before bootstrap. It does not change product hook behavior. Repaired recovery still runs through the real compact source. A correct stop is an expected terminal failure of the negative CLI invocation, whereas the positive invocation must complete normally. No timeout is accepted as completed repair.
+- [x] Complete source-scoped live C01–C16 evidence committed for current full run #27
+- [x] Finite C08 repaired-path completion verified live and committed for production closure
+
+## 7. Source-scoped qualification and release closure — reconciled 2026-09-08
+
+### Current qualification
+
+The current reviewed full result is selected by `qualifications/index.json.current_run`.
+These fields mirror its committed `qualification-summary.json`; the source is
+the actually executed commit, not the later evidence-import commit.
+
+| Field | Recorded value |
+|---|---|
+| Run | `34140846679` |
+| Source commit | `a9cdcdc1e0cad70e88b60869e75e4166046dd306` |
+| Live date | `2026-09-07` |
+| Codex | `codex-cli 0.153.4` |
+| Model | `gpt-5.6-sol` |
+| OS | `Debian GNU/Linux 13 (trixie)` |
+| Full release gate | `true` |
+| Finite C08 closure | `REPRODUCED` |
+
+Full run #27 reproduced all C01–C16, including the finite C08 stop/repair
+protocol without timeout. Its exact archive, summary and provenance are committed
+under `qualifications/34140846679/`; current capability packages are exact
+archive copies. C04 remains informational and fifteen capabilities gate release.
+
+C08's missing-checkpoint invocation stopped as intended with exit 1. After
+checkpoint repair, the positive invocation completed with exit 0; all nine
+positive protocol checks passed. C09 completed three reconciliations and two
+ordered automatic recovery cycles. Both independent C10 context-delivery probes
+completed with exact recovery values. See `QUALIFICATION_STATUS.md` for details.
+
+C10 delivers recovery through `SessionStart(source=compact)`; `PostCompact`
+remains advisory. C13's real project-scoped agent and hook passed via the
+explicitly allowed, known-error-gated non-ephemeral retry. Ephemeral custom-agent
+spawning is not claimed as working. Qualification applies only to the recorded
+Codex/model/OS configuration and does not guarantee future model behavior.
+
+### Preserved history
+
+Run #25 (`34060321283`, source
+`d0384f76bc4150d33bb8f51ef5981f3243b3cfb3`) passed the older narrow assertions,
+but its positive C08 trial timed out after proving unblocking. Its original
+archive, source identifier, hashes and limitation remain unchanged. It is
+historical evidence, not the current finite C08 publication proof.
+
+The replacement C08 fixture isolates the deliberate invalid-checkpoint trigger
+from ordinary startup recovery by narrowing SessionStart to `^compact$` in the
+disposable root checkout before bootstrap. It does not change product hook
+behavior. Full #27 verified this finite protocol live; no timeout is accepted
+as completed repair.
+
+### Publication and follow-up
+
+The finite C08 evidence prerequisite is satisfied. Production publication still
+requires strict validation on a clean tree, green protected CI, and a
+GitHub-verified signed annotated tag under `RELEASE.md`. A documentation
+reconciliation does not itself publish a release or require another live run
+when the qualified product/runtime inputs remain unchanged.
+
+[Issue #37](https://github.com/KeyffMS/PlanAnvil/issues/37) tracks intermittent
+receipt observation in hosted C09 loopback conformance. It remains open; a
+passing rerun does not establish the cause or fix the observer. The independent
+full #27 evidence remains valid and immutable. C13's retained transport
+limitation and its baseline 2.3 resolution are tracked in
+[issue #17](https://github.com/KeyffMS/PlanAnvil/issues/17).
diff --git a/docs/RELEASE.md b/docs/RELEASE.md
@@ -74,7 +74,7 @@ the full index, edit old actual observations, or label the denied automated laun
 as tested. Validate archive integrity and production readiness on the resulting
 clean tree through a protected PR before any signed publication.
 
-The active main ruleset was verified on 2026-09-07: PR-only squash changes,
+The active main ruleset was reverified on 2026-09-08: PR-only squash changes,
 seven required status checks, strict up-to-date branch, conversation resolution,
 no deletion or force push, no bypass actors. The protected distribution job now
 also requires real-CLI conformance to succeed. No policy is loosened for closure.
@@ -84,3 +84,25 @@ the evidence-import commit, qualification preparation is complete; a signed
 annotated production tag remains a separate authorized publication. This import
 does not create a tag or release. No unsigned or lightweight tag may substitute
 for the required verified signature.
+
+## Remaining publication steps and follow-up
+
+All changes intended for the first 0.2.0 publication are grouped under its
+pending-publication section in `CHANGELOG.md`. `VERSION` remains 0.2.0;
+qualification reconciliation alone does not change the distribution version.
+
+Before tagging, merge the documentation reconciliation through protected CI,
+run strict release validation on the resulting clean checkout, and use the
+verified signed annotated-tag procedure above. The current qualification
+metadata and C08 compliance checkbox are compared with the archived summary
+and index by `tests/test_qualification_documentation.py` in PR and release CI.
+
+The baseline 2.3 contract and full #27 evidence resolve the C13 qualification
+question in [issue #17](https://github.com/KeyffMS/PlanAnvil/issues/17).
+Ephemeral custom-agent spawning remains a documented compatibility limitation;
+requalify it when the upstream runtime changes.
+
+[Issue #37](https://github.com/KeyffMS/PlanAnvil/issues/37) remains open for
+hosted C09 receipt-observer reliability. It does not invalidate the separate
+full #27 proof. Future Codex/model/OS support needs scoped requalification.
+Plugin packaging remains outside v1 under `IMPLEMENTATION_SPEC.md`.
diff --git a/tests/test_qualification_documentation.py b/tests/test_qualification_documentation.py
@@ -0,0 +1,54 @@
+from __future__ import annotations
+
+import json
+import re
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+class QualificationDocumentationTests(unittest.TestCase):
+    def setUp(self):
+        self.index = json.loads((ROOT / 'qualifications/index.json').read_text(encoding='utf-8'))
+        folder = ROOT / 'qualifications' / self.index['current_run']
+        self.summary = json.loads((folder / 'qualification-summary.json').read_text(encoding='utf-8'))
+        self.compliance = (ROOT / 'docs/OPENAI_COMPLIANCE.md').read_text(encoding='utf-8')
+
+    def test_current_compliance_record_matches_archived_qualification(self):
+        # Scope to the current record so preserved historical runs cannot satisfy it.
+        section = self.compliance.split('### Current qualification\n', 1)[1]
+        section = section.split('\n### ', 1)[0]
+        rows = re.findall(r'^\| ([^|]+?) \| `([^`]+)` \|$', section, re.MULTILINE)
+        self.assertEqual(len(rows), len(dict(rows)), 'duplicate qualification fields')
+        self.assertEqual(dict(rows), {
+            'Run': self.index['current_run'],
+            'Source commit': self.summary['source_commit'],
+            'Live date': self.summary['date'],
+            'Codex': self.summary['codex_version'],
+            'Model': self.summary['model'],
+            'OS': self.summary['os'],
+            'Full release gate': json.dumps(self.summary['release_gate_passed']),
+            'Finite C08 closure': self.index['c08_closure'],
+        })
+        self.assertEqual(self.summary['github_actions_run'], self.index['current_run'])
+
+    def test_finite_c08_checklist_matches_committed_closure(self):
+        matches = re.findall(
+            r'^- \[([ x])\] Finite C08 repaired-path completion verified live and committed for production closure$',
+            self.compliance, re.MULTILINE,
+        )
+        self.assertEqual(len(matches), 1, 'finite C08 checklist must have one current entry')
+        self.assertEqual(matches[0] == 'x', self.index['c08_closure'] == 'REPRODUCED')
+
+    def test_current_version_notes_include_current_source_bound_result(self):
+        version = (ROOT / 'VERSION').read_text(encoding='utf-8').strip()
+        changelog = (ROOT / 'CHANGELOG.md').read_text(encoding='utf-8')
+        section = changelog.split(f'## [{version}]', 1)[1].split('\n## [', 1)[0]
+        for value in (self.index['current_run'], self.summary['source_commit']):
+            with self.subTest(value=value):
+                self.assertIn(f'`{value}`', section)
+
+
+if __name__ == '__main__':
+    unittest.main()