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.

V6.2 Release Audit Report

Audit date: 2026-08-28

Scope

  • Whether V6.1 field repairs were fully incorporated into V6.2 rather than only changing the version number.
  • Native Windows Codex 0.150.1 Plugin and Marketplace registration.
  • Installation, backup, rollback, uninstall, protection of unknown external files, and long-path behavior.
  • Ten Skills, seven Reviewers, six Hooks, and TaskOutcomeEvent V2.
  • Model-cost ceiling, self-observation isolation, and controlled-evolution authorization boundaries.
  • Neutral natural language and compatibility boundaries for machine contracts.

Conclusion

Passed. V6.2 formalizes the V6.1 field repairs and adds Windows extended-length path support with regression tests. A real isolated environment completed the V6.1 -> V6.2 -> V6.1 closed loop, and Codex CLI readback confirmed the target Plugin's installation, enabled state, and version.

Natural-language content uses neutral wording and test identities use anonymous placeholders. Machine contracts such as --scope user, JSON field names, and path variables remain unchanged for compatibility.

Key Assertions

  • Plugin ID: codex-cross-project-engineering-assistant@cp-assistant-local.
  • Marketplace: cp-assistant-local.
  • V6.2 success requires installed=true, enabled=true, and version=6.2.0.
  • Automatic subagents allow only four Luna/Terra tiers with a maximum of gpt-5.6-terra + high.
  • Reviewer TOML does not fix models; this package does not override the main-agent model.
  • SessionEnd timeout is three seconds; Windows Hooks do not require python3.exe.
  • Proposal always keeps execution_authorization=NONE; human ACCEPT is not execution authorization.
  • Installation failure rolls back the managed-target list. Uninstall merges and restores managed AGENTS and standalone Hook content without deleting unknown account Skills, Agents, Hooks, MCP configuration, other configuration, or historical observations.

Evidence Index

  • VALIDATION_REPORT_V6.2.md
  • RELEASE_NOTES_V6.2.md
  • V6.2_BUILD_INFO.json
  • tests/test_package_manager_security.py
  • tests/test_v60_deterministic_observation.py
  • scripts/package_manager.py
  • scripts/validate-package.py