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.

V6.6.1 Release Notes

Chinese: Chinese documentation

Delivered

  • Two reproducible, independent archives: zh-CN and en.
  • Full English coverage for every natural-language document, all ten Skills and their References and templates, seven Reviewer definitions, examples, structured descriptions, and Python runtime messages.
  • Paired Chinese/English source comments and docstrings, plus a fail-closed build-time catalog of human-reviewed English runtime literals.
  • Bounded retry for transient Windows atomic-publication sharing failures.
  • Explicit validation-only worker waiting, while production SessionEnd remains detached and bounded by the three-second host timeout.
  • Exact Plugin acceptance remains installed=true, enabled=true, and version=6.6.1.

Unchanged safety policy

  • execution_authorization=NONE
  • no automatic Skill or Reviewer modification
  • no automatic proposal acceptance or execution
  • no automatic commit, push, deployment, restart, production operation, or data write
  • no hard-coded Reviewer model
  • automatic model ceiling remains gpt-5.6-terra + high

Runtime evidence statement

Codex 0.150.1 still supplies diagnostic observations rather than trusted runtime model attestation:

requested_model_policy = PASS
runtime_model_evidence = UNAVAILABLE
diagnostic_model_observation = diagnostic only