ScopeProof v0.2.6 — downloadable evaluation release

ScopeProof v0.2.6 is a downloadable evaluation release for explaining a saved acceptance decision, sharing it offline, and safely revisiting it after changes. It remains a local, reviewer-controlled evidence assistant for public GitHub pull requests.

What changed since v0.2.5

• Decision overview after analysis or reopening: reviewed head, verdict, blockers, unresolved file scope, recorded final acceptance and save status.
• Gate-reason differences between saved analyses, including cases where the verdict stays the same.
• Offline HTML report download and optional rationale for recorded final acceptance or revocation.
• Saved requirements prepared as fresh unconfirmed drafts, compared across analyses and exported with provenance.
• Clearer source-conflict and refused-comparison explanations; pending scope drafts and stale deletion consent remain protected.
• Windows concurrent-save locking and constructed browser-check maintenance.

Source and downloads

Source commit: 68f5c7af9cdaafe97ab72c1a7f6c8784f65a44e1. The annotated v0.2.6 tag and release target must identify this same commit.
Package version: 0.2.6; ruleset generation: 2.0.0; saved-record format: v9.
Core Python source fingerprint: 7b3a9180fa8658e800486cac4cc0cb1a3ff236bb88c0ad9c9f165a699e04eb64. This excludes dependencies, UI and platform; artifact hashes below identify the full uploaded bytes.

Uploaded asset	Bytes	SHA-256
scopeproof-0.2.6-py3-none-any.whl	448,322	48b308b3f92d912932734aa4907e958d78351bb956f1a9ce65353e9206741896
scopeproof-0.2.6.tar.gz	7,746,208	8088ee5d1b8bd38eabfb245b52b860fc89219c47c9cfc78ea7d1110c4c239e65
SHA256SUMS.txt	190	b6f6f3833ce0479812eb42994dfc5aec2ea23907ec3f9ed55a3085ae01cd3252

Use the named wheel or source archive with SHA256SUMS.txt; GitHub-generated source ZIP/tar links are separate from the verified package assets. Follow the versioned installation guide and evaluation guide. Only final-version artifacts were built; the earlier dev0 candidate was not renamed.

Automated engineering verification

• Independent automated source review approved PR #271 head 936ecb85309c98f6fe52372a557c55a14f65e724; PR CI and resulting-main CI passed on their own exact heads.
• Resulting-main suite: 4,842 passed, 36 skipped, 95.5067% coverage; 7 installed CLI checks and 32 packaged-browser collection checks. Python 3.11/3.13, Windows compatibility and locked-environment lanes passed.
• Three clean final-wheel/source baseline installations on macOS Python 3.11/3.13 passed package identity, dependencies, 12-case/13-criterion acceptance and two-case comparison benchmarks, with zero mismatches.
• Seven exact-wheel CLI journeys, seven full Chromium journeys and both installed launcher health checks passed. These journey results apply to merged-main artifacts because the rebuilt final wheel and source archive are byte-identical to their verified previews.
• Two constructed native v9 histories written by the actual public v0.2.5 package survived candidate read, copied resave, transfer, recipient reopen and report provenance checks. Original bytes, producer metadata and complete state/history remained intact; pending criteria stayed unconfirmed.

Public download verification

Public download verification: completed automatically on November 7, 2026.

• Five anonymous HTTP checks returned 200 with curl configuration disabled and no supplied authentication: release page, latest-release redirect and all three uploaded assets. Latest resolves to v0.2.6.
• Downloaded wheel, source archive and checksum manifest are byte-identical to the verified final assets above. Both manifest entries and package metadata version 0.2.6 were checked.
• Fresh public-wheel and public-source installations on macOS Python 3.11.15 passed installed module/file identity, locked dependency checks, the 12-case/13-criterion acceptance benchmark and two-case comparison benchmark, with zero mismatches.
• Both installed workbenches returned loopback health ok and shut down cleanly. The downloaded wheel passed seven installed CLI checks and two full Chromium journeys (primary workflow and saved-review comparison), with no failures or skips. No related processes remained.
• The complete Pydantic verification record was revalidated and its 118 nested evidence-file hashes checked. Prepublication native-upgrade evidence applies through identical final wheel bytes; a new public native-upgrade run is not claimed.

These are constructed automated engineering checks and contribute zero genuine-use or qualifying-review credit.

Evaluation scope

Back up saved reviews and preserve original files before switching versions; downgrade support is not promised. Compatibility observations cover constructed records, not every possible history.

The static matcher can miss relevant evidence. Candidate implementation or test references do not prove target runtime behavior or correctness. Normalized criteria require explicit confirmation, decisions remain attributable reviewer records, and target-repository code is never executed by the application server. Runtime values in constructed fixtures are assertions for engineering tests, not observed target execution.

The evaluation-only use policy, public-only intake, local platform limits and asserted identities remain in force. Browser journeys are POSIX automation; Windows CI is not a Windows desktop usability certification. Frozen R-002 inputs and historical results are unchanged.

All tests, automated agent reviews and this release are engineering evidence. No personal testing or qualifying reviews were required. Genuine independent use, customer validation, accessibility conformance and a completed beta are not claimed. Stage 3 remains dormant; this release does not deploy a hosted app or initiate outreach.