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.

V7.0.0 Release Notes

Chinese: Chinese documentation

Version: 7.0.0

Main changes

  • Added language-neutral $backend-engineering for Node.js, Go, .NET, Rust, and mixed-language services, with Java and Python guidance loaded on demand.
  • Added independent $ai-engineering for model integration, structured output, RAG, agents, evaluation, inference, and multimodal systems.
  • Kept $frontend-engineering independent; $data-middleware-infrastructure now focuses on databases, middleware, storage, GPU resources, containers, and network runtime boundaries.
  • The former Java, Python+AI, and Data+AI Skill names are no longer routable Skills; upgrades remove only the four Manifest-declared legacy directories, including the previously deprecated Vue Skill.
  • Synchronized the bilingual responsibility matrix, positive/negative routing cases, Manifest, AGENTS, recovery guide, and release tooling with 7.0.0.
  • Added independent Chinese and English documentation sites, repository-wide Markdown link auditing, versioned release evidence, reproducible bilingual artifacts, and a draft-only GitHub Release provenance workflow.

Unchanged safety boundaries

  • execution_authorization=NONE
  • Skill activation does not expand file, Git, environment, production, or data authority
  • No automatic commit, push, publication, deployment, restart, or production write
  • Automatic sub-agent ceiling remains gpt-5.6-terra + high

Acceptance boundary

Package validation and the real Codex implicit-trigger observation are recorded separately. Four representative implicit-routing scenarios, a source-tree 6.6.0 -> 7.0.0 Plugin upgrade, and fresh-task routing PASS. The public Release ZIP still requires independent provenance and post-download acceptance; this does not establish deterministic selection for every prompt variation.