ScopeProof v0.2.1 — Re-review evidence integrity

ScopeProof v0.2.1 makes re-review evidence changes explicit and aligns the public release path with the implemented product.

Highlights:

• classifies re-review candidates as Unchanged, Relocated, Modified, Added, or Removed with conservative ambiguity handling
• provides two-sided JSON, Markdown, and Streamlit comparison output anchored to immutable evidence references
• includes the checked-in scopeproof comparison-benchmark engineering benchmark
• aligns package version, install URLs, Pages CTA, and commercial research documentation
• replaces stale direct-message outreach instructions with the canonical inbound-only public-alpha issue form

Evidence boundary:
The comparison corpus is deliberately constructed engineering evidence. It does not advance Stage 1, prove correctness, establish external use, or constitute customer or market validation. No paid API, LLM, billing, private repository, account, fork test, email, direct message, automated outreach, scraping, synthetic validation, generic review, security scanner, or automatic fix capability was added.

Verification:
Protected PR #158 passed verify, Python 3.11 compatibility, locked-environment, and CodeQL checks. Main CI, CodeQL, and Pages also passed on merge SHA 80a3ff946dbd30efbc8058883b89f73d9486d0ee. The attached wheel, source archive, and checksum manifest were rebuilt from that exact SHA.

ScopeProof v0.2.0 — Reviewer-first product reset

ScopeProof v0.2.0 resets the product around deterministic acceptance-coverage review.

Highlights:

• Human reviewer language for evidence candidates and review status
• Bounded evidence context with immutable matched lines
• Atomic external verification and deterministic final-acceptance prerequisites
• Explicit bounded candidate paths and immutable re-review comparison
• Optional, consent-separated alpha feedback flow
• Observed CI semantics that count only explicit success

Verification:

• 804 tests passed, 1 skipped, 96.28% coverage
• Ruff passed
• 12 benchmark cases / 13 criteria, zero mismatches, zero must-have False Ready, zero incorrect citations
• Fresh wheel install verified for scopeproof and scopeproof-web 0.2.0
• Main CI, CodeQL, and Pages passed at d84ea24

Boundary: this release does not claim genuine participant validation or beta readiness. Current alpha state remains waiting_for_external_participant_evidence. No paid API, billing, private repository, fork testing, untrusted-code execution, generic review, security scanning, or auto-fix capability is included.

ScopeProof v0.1.23

ScopeProof v0.1.23

This corrective public-alpha release is the first wheel that contains the participant workflow
documented by ScopeProof's ten-minute quickstart.

Included

• scopeproof alpha init for explicit source-owner and public-information qualification
• scopeproof alpha outcome for one bounded participant outcome
• scopeproof alpha show with consent-gated public summaries and no identity fields
• guided workbench onboarding that separates confirmed alpha cases from technical smokes
• reproducible Python 3.12 environment lock and Python 3.11 compatibility coverage
• public GitHub Pages site and deterministic, captioned 60-second constructed demo
• copyable GitHub Action pinned to the immutable v0.1.23 package tree

Verification

• 769 tests passed and 1 live-network test was skipped
• 96.80% local coverage against a 95% requirement
• 12 deterministic benchmark cases had zero mismatches and zero must-have False Ready outcomes
• archive-built wheel, checksum, dependency health, installed alpha CLI, policy provenance, and
  loopback workbench health were verified
• protected CI and CodeQL passed before merge and on the release commit

These are controlled engineering checks. They are not external-user validation, runtime evidence
for a participant's pull request, adoption, or proof of correctness. ScopeProof remains a
local-first evidence assistant; it does not execute untrusted repository code or replace QA.

No paid API, billing, email, automated outreach, fork testing, private repository, or synthetic
evidence is included.

ScopeProof v0.1.22

ScopeProof v0.1.22

This consolidated public-alpha release contains the protected improvements accumulated since v0.1.21:

• verifies the declared Python 3.11 minimum alongside the primary Python 3.12 path;
• documents the complete Markdown, JSON, CSV, and HTML export inventory;
• adds evidence-gated public-alpha, limited-beta, and beta-expansion governance;
• adds a disclosed constructed-demo onboarding visual;
• records the repository's evaluation-only use policy without granting an open-source license;
• carries exact evaluation-policy bytes and the canonical Use Policy URL in the wheel;
• avoids duplicate full CI runs for feature pull-request SHAs while retaining an independent post-merge main run.

Controlled verification

• 723 tests passed with one intentional live-network skip.
• Ruff and dependency consistency checks passed.
• The deterministic benchmark executed 12 cases and 13 criteria with zero mismatches, zero false blockers, and zero must-have False Ready outcomes.
• Python 3.11 compatibility, protected verify, and CodeQL passed before merge.
• The release wheel was built from protected-main commit b00a2aae4d46178d49b447619b342902baa49aa0.
• A clean installation confirmed distribution, module, CLI, web launcher, and new-review identity as 0.1.22.
• Installed policy bytes matched the repository policy; no open-source license metadata was declared.
• The installed local workbench health endpoint returned ok.

These are controlled engineering, packaging, and local health checks. They are not external pull-request runtime evidence, customer validation, adoption evidence, or proof of correctness.

No paid or LLM API, billing, organization, private repository, fork test, generic code review, security-scanning product feature, or automatic fix capability is included.

ScopeProof v0.1.21

ScopeProof v0.1.21 makes the safe local scopeproof list command available through the verified public wheel and aligns the README installation contract with installed behavior.

Controlled verification: Ruff passed; 710 tests passed with one intentional live-network skip; the 12-case/13-criterion deterministic benchmark reported zero mismatches, must-have False Ready outcomes, and false blockers; fresh wheel installation, checksum, dependency, CLI identity, local-list behavior, and loopback workbench health checks passed.

These controlled repository and package checks are not external pull-request runtime evidence, customer validation, or proof of correctness. ScopeProof remains an evidence assistant and requires user-confirmed criteria and human acceptance.

ScopeProof v0.1.20

ScopeProof v0.1.20 consolidates 12 protected public-alpha improvements since v0.1.19.

Highlights:

• clearer public-PR fetch recovery and review availability
• safer criteria, requirements, and criterion-detail draft persistence/export boundaries
• complete manual runtime-evidence records with persisted UTC timestamps
• attributable resolution history and explicit decision reviewer identity
• immutable copyable GitHub Action source pinned to the exact v0.1.20 release tree

Controlled verification: Ruff passed; 703 tests passed with one intentional live-network skip; the 12-case/13-criterion deterministic benchmark reports zero must-have False Ready, zero false blockers, and no mismatches; clean wheel installation, dependency checks, both CLIs, new-review provenance, and loopback workbench health passed.

Evidence boundary: these are controlled product and packaging checks. They are not external PR runtime evidence, customer validation, or proof of correctness. Fork testing remains permanently excluded. No paid or LLM API is used.

ScopeProof v0.1.19

ScopeProof v0.1.19 consolidates the installed workbench improvements merged in PRs #110–#122.

Highlights:

• clearer confirmed-analysis, source-identity, storage, runtime-evidence, criteria, and final-acceptance recovery guidance
• sidebar links for stable workflow sections without false navigation affordances
• immutable GitHub Action install pin to the exact v0.1.19 release package tree
• checksum-verifiable wheel install instructions

Verified on exact protected main b6d09df:

• Ruff passed
• 675 tests passed; 1 intentional live test skipped
• deterministic benchmark: 12 cases / 13 criteria, 0 mismatches, 0 must-have False Ready, 0 false blockers
• fresh wheel installation, dependency check, CLI/web/new-review version identity, and loopback health passed

Evidence boundary: these controlled tests, fixtures, and local runtime checks validate ScopeProof product behavior and packaging. They do not prove that any external pull request meets its acceptance criteria and are not external runtime evidence. No forks, billing, paid APIs, organizations, private repositories, synthetic users, or synthetic validation were used.

ScopeProof v0.1.18

ScopeProof v0.1.18

ScopeProof v0.1.18 packages 29 protected improvements merged after v0.1.17.

Highlights

• Revalidates review, criteria-revision, runtime-evidence, resolution-event, lifecycle, and export inputs at Pydantic trust boundaries.
• Binds analysis and resolution history to the correct review and active criteria revision, preserving audit lineage across reanalysis.
• Enforces deterministic gate inputs and rejects invalid or blank evidence context instead of risking False Ready.
• Preserves ingestion limitations and renders evidence references and runtime guidance safely and durably.
• Requires explicit per-PR Action applicability and reviewer-confirmed requirements while keeping required-check mode disabled.
• Adds confirmed single-record local review deletion in both CLI and workbench, with explicit export and secure-erasure boundaries.

Controlled verification

• Protected merge commit: c52e87340d33bab49c776635e2dd3a328b9074b2.
• Ruff passed; pytest reported 658 passed and 1 skipped.
• Deterministic benchmark: 12 cases, 13 criteria, 0 mismatches, 0 must-have False Ready, 0 false blockers, 0 evidence-link errors, and 0 unexecuted categories.
• Fresh installed-wheel checks passed for dependencies, distribution metadata, module version, both CLIs, new-review provenance, benchmark, and loopback workbench health.
• Public full-SHA Action source installation and benchmark passed.
• Wheel SHA-256: 32053552fe976b39a2afc86f67707882b8a388e558dbd0da2d4449cad05c19e3.

These results are controlled source, packaging, benchmark, and loopback-health evidence. They are not external PR runtime evidence, customer validation, or proof of correctness. This release uses no billing, paid API or LLM, private repository, organization, second account, or fork test.

ScopeProof v0.1.17

ScopeProof v0.1.17

ScopeProof v0.1.17 packages 13 protected improvements merged after v0.1.16.

Highlights

• Clearer prepared-criteria handoff before analysis.
• Strict rejection of blank acceptance criteria, runtime context, runtime limitations, Action evidence context, and note-free manual verification.
• Stronger GitHub Action evidence identity checks for repository, owner, commit SHA, and event head SHA.
• Discoverable saved local reviews and safer local review-store root validation.
• Copyable GitHub Action source pinned to the immutable v0.1.17 release tree.

Controlled verification

• Protected merge commit: f62590bc2ad8c22cb2d9a57d3b245d85bbde9da1.
• Ruff passed; pytest reported 353 passed and 1 skipped.
• Deterministic benchmark: 12 cases, 13 criteria, 0 mismatches, 0 must-have false-ready, and 0 false blockers.
• Fresh installed-wheel checks passed for dependencies, distribution metadata, module version, both CLIs, new-review provenance, benchmark, and loopback workbench health.
• Wheel SHA-256: e0ef44d78219cb99a2b8b8f240c25323ec271e37e6701044d129e41176c3d4c3.

These results are controlled source, packaging, benchmark, and loopback-health evidence. They are not external PR runtime evidence or customer validation. This release uses no billing, paid API or LLM, private repository, organization, second account, or fork test.

ScopeProof v0.1.16

ScopeProof v0.1.16

ScopeProof v0.1.16 consolidates six protected improvements to the local public-PR review workflow:

• Evidence-level guidance now appears where reviewers confirm criteria and inspect observed evidence.
• Gate reason codes are rendered as readable product language without changing deterministic gate data.
• Criteria confirmation uses one durable success state.
• The evidence matrix is reduced to scan-oriented fields; confidence, candidate count, reason, and missing evidence remain in criterion detail.
• Saved-review reopening is a collapsed secondary path so first-use actions remain primary.
• Visible criterion edits are explicitly marked as pending confirmation. ScopeProof preserves and labels the last confirmed analysis, prevents another analysis run across the mismatch, and invalidates the old analysis when the updated criteria are confirmed.

Verified release identity

• Protected-main merge: e6f2ad46c7cf56f7ce2886d20d6b2a8aa7f30c22
• Immutable Action source tree: f567447b4d23ef7e81df789b48826c221cb8a81e
• Offline source suite: 243 passed, 1 intentional live-network skip
• Deterministic benchmark: 12 cases, 13 criteria, 0 mismatches, 0 must-have False Ready, 0 false blockers
• Wheel: scopeproof-0.1.16-py3-none-any.whl
• SHA-256: 64f0145d6f89229e62cfbebe8cef18fc996e2ab8a81b1b6ab8a234e5c5851088
• Clean installed identity: distribution, module, CLI, web launcher, and new-review provenance all report 0.1.16
• Dependency check: no broken requirements
• Packaged workbench: exact loopback health response ok on port 8517, clean shutdown, no traceback
• Protected PR checks, merged-main CI, and merged-main CodeQL succeeded before publication

The bundled demo, automated suite, benchmark, packaging checks, and local health smoke are controlled product verification. They are not runtime evidence for a pull request, external customer validation, or proof of correctness.

Boundaries

ScopeProof remains deterministic, local-first, public-repository-only, and informational. It uses no paid or LLM API, creates no billing dependency, does not execute pull-request code, and does not support fork testing, private repositories, organizations, or second accounts in this public alpha.