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.

V7.13.6 Validation Record

  • All 57 targeted tests passed for implementation commit e74c6b9, covering exact V1 names, neutral unbound behavior, bound reentry denial, and unchanged accounting.
  • The initial full validation had one session-end concurrent-recovery failure. Twenty repeated runs of that case and all 20 tests in its module passed. The original cause remains unconfirmed; its evidence is retained without changing assertions or timeouts. Version metadata, documentation, artifacts, and the remote matrix are verified separately.
  • A real V1 control on local Desktop bundled component 0.158.0-alpha.2.1 returned V4_FRESH_DISPATCH_REQUIRED. The single child actually ran GPT-6 Luna / Low without tools; native receipt and completion records accounted for one attempt/unit.
  • Positive admission proves the tested budget transport chain only, not model quality, other model/effort tuples, or the V2 encrypted-message channel. GPT-6 production defaults remain disabled.
  • Patch artifacts, installed digests, CI, signatures, and public assets are read back independently.