跳转至

Historical version

This page records design, operation, or validation facts from an earlier release. It is not current V7.17.3 guidance. Start from the documentation hub for current instructions.

V7.8.0 Validation Record

This record reports source-candidate, public-release, and active-account effectiveness separately. Final counts are read back only after the release commit stabilizes.

Boundary Contract
C01-C25 The static registry contains exactly 25 entries. An unclassified Manifest, Hook, or CLI entry blocks the all-covered claim without blocking ordinary conversation.
AUTO and authority OFF, maximum, prerequisites, and risk are computed independently; authorization=false and action_status=NOT_AUTHORIZED are fixed.
First use Only readback-confirmed ACCEPTED_PERSISTED can queue a full scan; no reply, decline, or write failure keeps BASIC available.
Scan jobs Offer -> scan -> index lock order, lease generation, cancel epoch, index commit before cursor, late fencing, and three-attempt takeover ceiling.
Installation and recovery Actual Python 3.11+ probes, feature doctor, read-only inventory, canonical recover, zero-delete state-less preview, and non-Git BASIC guidance.
Review U02-U13, UX01-UX40, M01-M12, and T25-T28 have exact contracts and discoverable tests; postimplementation review is separate evidence.

Local candidate evidence

  • After integrating the independently reviewed and released V7.7.1 / Codex 0.154.0 compatibility fix and correcting Windows path aliases, a full python scripts/validate-package.py rerun passed 422 package + 205 runtime tests on Python 3.13.15.
  • Focused results: capability registry 10, onboarding 15, scoped review 4, installer UX 13, installer security 40, and R3 traceability 76, all PASS. Three path-containment regressions also passed on the official portable Python 3.11.9 runtime.
  • Windows lock-file concurrent initialization passed 30 consecutive stress iterations. Documentation, localization, links, semantic, privacy, routing, delegation, and payload gates passed.
  • Fifty real disabled/unconfigured file-gate processes measured p50 82.22 ms, p95 87.35 ms, and p99/max 88.46 ms, within the p95 <= 300 ms and p99 <= 1 s budgets.
  • R3 postimplementation state/concurrency and test/delivery Reviewers passed premerge refreshed packet f22009...a01402. The decline-cancel compensation and false C20 entrypoint HIGH findings were repaired. V7.7.1 compatibility changes have their own independent review and release evidence. The combined baseline received the full local rerun above and is not represented as a new independent Reviewer pass.
  • Public CI, Release, anonymous download, account installation, restart, and fresh-task loading are not yet established by this record.

Separate statuses

  • RELEASE_COMPLETE requires commit, PR merge, tag, release workflow, public assets, anonymous downloads, and isolated installation readback.
  • INCIDENT_EFFECTIVE requires active-account installation, any necessary restart, fresh-task loading, and original-project message acceptance. It is confirmed separately from RELEASE_COMPLETE.
  • This file enters source as a candidate record; no project may claim either state without subsequent evidence.