v3.34.0 — Evidence that reaches its subject

Theme: Evidence that reaches its subject.

What's New

• Receipt continuity distinguishes unrelated changes from changes to an accepted target.
• Payload lint and capability-retention checks expose inherited backlog and destructive removal.
• Canonical discovery and template scaffolds improve the shipped study and instantiation paths.
• Receiver, host-layout and runtime checks report unavailable evidence explicitly.
• Voice-consumer and standards bindings connect declared inputs with the evidence inspected.
• Core and thirteen registered templates carry the release payload and current version declarations.

Compatibility

Python 3.10 or newer is required. Preserve receiver-owned extensions and inspect local changes before upgrading.
Follow the tag-bound release handoff; publication does not establish installation or receiver acceptance.

Known Issues (Pre-existing)

This release explicitly carries measured citation, assurance-traceability, dependency-reference, skill-baseline, vocabulary and handoff limitations. Raw failures remain failures; missing evidence remains unavailable.
The framework manager owns carried findings at the first post-release maintenance intake, before next-version scope. No blanket correctness claim is made.

Disclosed Limitations

Historical regression-coverage shortfalls remain disclosed; accepted implementation repairs do not establish complete release or downstream coverage. See the full release notes and exact limitations and CHANGELOG.

Known Gaps

Independent partition review retained failures in derived manifest versions, version-carrier detection, predecessor enforcement, and release-body validation. These remain failed measurements, not repaired capabilities.

Further packaging and documentation findings include stale core metadata, template migration histories missing the current entry, six templates without a license file, an absent version-specific delta document, and gaps in the release plan's validation instructions. Some historical classifications were established only against the pre-theme candidate, rather than the previous public release; this distinction remains unresolved and is not presented as proof that every finding is inherited.

The review did not establish receiver breakage from these findings; they are disclosed under the release-specific publication exception. The framework manager owns their disposition at the first post-release maintenance intake. Independent review does not establish full requirement compliance, complete platform execution, installation or receiver acceptance.

Handoff and remote migration-message summaries and tracking structure remain incomplete; the document-processor template lacks a root manifest, and the core changelog's Unreleased section is misplaced. Platform-hazard documentation and execution coverage remain explicitly limited. These results remain disclosed, not repaired by the theme additions.

Independent Assessment Coverage

Three serial independent assessments returned findings and separate coverage ledgers for all 437 registered source rows. Those rows overlap and are not 437 distinct gates. Results: 35 failed, 264 unavailable, 94 unassessed, 17 partial, 25 narrowly scoped passes, one not applicable and one withdrawn. Complete obligation coverage and full independent release acceptance were not established. Publication proceeds under an explicit release-specific exception retaining those failures and coverage gaps; no introduced receiver-breaking defect was demonstrated in the inspected scope. Actual protected-branch checks and public-source verification are separate publication evidence. Following those assessments, one focused independent review accepted a two-file CI assembly correction: the test jobs install Ruff 0.16.7, and three synthetic fixture paths use an already-recognized placeholder. Unaffected assessment evidence is reused; its coverage limits stand. All seven required checks passed on the resulting exact core candidate. Python 3.10–3.14 each installed Ruff 0.16.7 and returned 899 passed and 85 skipped tests, plus 27 focused contract passes. The previous failed run remains a failed historical result.

Post-Tag Repairs

• 2026-09-23: publication metadata (#106). codemeta.json and CITATION.cff named version 3.33.1 at this tag; they now name 3.34.0, dated 2026-09-14. The CHANGELOG [Unreleased] section, stranded below [3.30.0] since v3.30, is back at the top with its pointer on 3.34.0. No code, spec or test changes. The released package contents are unaffected.