Dated recovery history; closeout complete. PR #87 documentation publication, production deployment and post-merge CI (156 / 156 / 289, zero failures/errors/skips) are complete. No recovery task remains. Earlier pending states below are historical. Pre-reset commands mentioned in this record are not exposed by the bounded-v1 entrypoint. Use the root README, docs index and current closeout. The rc.1 tag and five assets are unchanged.

Bounded v1 recovery — public validation closeout

v1.0.0-rc.1 is published as a prerelease, not Latest. Main acceptance and prerelease authorization are complete. The documentation on this branch is a separate later proposal; it is absent from the rc.1 source tree. The source still reports 1.0.0; use the tag/commit/manifest to identify rc.1.

Supported v1 release matrix

The three retained hosted runs each passed 156 Git-only tests on Python 3.11, 156 Git-only tests on Python 3.14, and 289 combined Muse tests on Python 3.14. Each has zero reported failures/errors/skips. The public evidence index binds every run to its candidate context. No unresolved supported-test failure remains, and this documentation action did not repeat a closed review or unchanged suite.

Default collection is tests/v1 plus tests/retained; adding tests/muse selects the combined matrix. The exact supported driver is tools/ci/supported_v1.py, run with the kit's conventional .venv/bin/python and required --junitxml PATH. Optional --muse-python must select the separately installed pinned Muse runtime. Git-only operation requires no Muse. See test scope.

Hosted evidence is Linux with Python 3.11/3.14. Retained local evidence is macOS arm64 and fixed-path source installations. It does not certify Windows, arbitrary interpreters, packaged/desktop installs, relocation or arbitrary old-version rollback.

Earlier reviews, installation and pilot

The bounded architecture review and completed-build review closed after correcting the duplicate-selector abbreviation and executable-mode sync findings. The earlier 77-test Git core, two clean dependency installations with six fixture lifecycle stages, and 32 checks in the separately authorized DINERO manual pilot are retained at their original scope. They are not additional tests in the hosted totals and are not new reviews of this documentation proposal. The pilot preserved consumer files and kept automatic hooks off. No consumer was accessed by this follow-up.

The frozen detailed validation ledger preserves earlier findings, commands, corrections and milestone results. Its local paths identify historical operator evidence; they are not downloadable links or installation instructions. The full later local ledger is also preserved by the operator. This portable closeout replaces repeated historical NEXT directions with current conclusions; it does not erase the retained records or their failures.

Historical baseline failures

Before the scope reset, the full old suite reported 1,391 passed / 104 failed; the filtered run reported 1,313 passed / 44 failed / 138 deselected. The full failures comprised 68 sandbox socket denials, 16 launcher/PyYAML failures, nine missing paths/fixtures and 11 other product/test failures. Those former-product results are not called green. All 104 failing node IDs and the historical test sources/inventory remain preserved. Deferred desktop, hosted, orchestration and old publishing behavior is outside the bounded CLI.

MuseHub restoration audit — 2026-10-08

The owner corrected the mistaken removal of Muse authority alongside OCI and removed the session-count ceiling. Git-only is a permanent choice; Muse adoption is explicit, and the kit itself publishes Muse-first. The original 18-finding audit records reproduced bridge defects, evidence classes and feature-by-feature scope. R1 supplied local Muse authority/adoption; R2 supplied controlled mirror preparation and retry; R3 reconciled source and validated combined installation/rollback. Later credential compatibility repair and actual delivery evidence are recorded below. The obsolete budget gate and pending restoration language are historical. OCI, native GSR, sole-writer machinery, automatic governance rewrites and packet chains remain excluded; the old deploy path must not be reactivated.

R2 controlled mirror preparation and retry — 2026-10-08

The controlled mirror workflow separates local prepare/verify from explicitly authorized delivery. It pins source, destination, expected heads, exact files/modes and plan bytes, refuses dirty/foreign targets, and never exports onto a development tree. Fixture results remain local evidence; the later live sequence is reported separately below.

Muse artifact and CI readiness — 2026-10-08

The optional runtime is an Overseer recovery build of Muse 0.2.1rc5, recovered from verified installed files. It is not an upstream release or reconstruction of the unavailable original sdist. The artifact contract, recovery recipe and unchanged companion guide retain all four digests and the immutable wheel/source ZIP URL. CI uses MUSE_RC5_RECOVERY_WHEEL_URL; no moving installer or alternate package is accepted.

Hosted CI execution — 2026-10-09

The relevant retained release runs are feature 37983390020, PR 37985479580 and accepted-main 37996198972, each 156 / 156 / 289 passed. Earlier candidate runs are historical; their results are not transferred to a different candidate. No new hosted run was dispatched for this local docs plan.

Local credential compatibility repair — 2026-10-09

The repaired helper accepts repeatable Git challenge metadata while retaining strict scalar destination fields, exact HTTPS repository scoping, bounded UTF-8 input and in-memory credentials. The old helper reproduced six fixture failures; all 39 credential tests passed after repair. Subsequent non-force Git delivery authenticated successfully. No credential reset or persistent helper change was needed. See the immutable released helper/tests and the publication record.

Repaired candidate live publication — 2026-10-09

The Muse publication record binds the frozen recovery Git, accepted Muse revision/snapshot, real two-parent mirror and exact GitHub acceptance. Native source readbacks verified 861 file bytes. A first mirror authority read failed before push; a matching read and same-plan retry succeeded. The original refs/edits, isolated Muse feature/main and held Git-only release 06f988e5a078ede81c9dc664520833980a9a19a3 remain preserved.

GitHub acceptance — owner confirmed

The owner accepted PR #86's squash merge to 293e79eb96dc9b94b76fa28ffad5e9e0a6e85b30. Its tree 735b2eb9248148e5c11d377fa79fb1f4dc11b75f matches mirror 9c9db93b47ceabd9a021a1d0679cb696d65cc6c2. The original two-parent mirror history remains intact; no corrective merge or history rewrite is requested.

Prerelease published — 2026-10-09

The release closeout records release ID 408458432, the exact tag, five checksum-bound assets, and independent draft and published download verification. Publication is complete. These later docs do not retag rc.1, replace its assets, or change source version metadata.

Declared limits

Native rc5 clone omitted 19 empty directories while all 861 file bytes matched. Only immutable-snapshot-approved metadata was restored in isolated readbacks. This limitation remains; dirty-state guards must not be bypassed. Muse snapshots lack POSIX modes; the ten executable paths require explicit policy.

Private/production Hub state and trust, native linked-worktree mutation/stage isolation, general file/mode fidelity, semantic-analysis breadth, relocation, untested platforms and broader consumer rollout remain outside the evidence. Recovered runtime provenance and rc.1/source-version distinction stay explicit. No stable or complete-product readiness follows from green bounded tests.