v4.0.1
Codex-primary compatibility release for ChatGPT-backed runtimes. Delegate configuration now fails early with actionable diagnostics, generated skills use each provider's native invocation syntax, and cross-provider review behavior stays observable.
Fixed
- Codex
soldelegates now resolve to the supportedgpt-5.6-solmodel ID; all Codex agent model, effort, and sandbox combinations are validated before generation. - Codex-generated skills render
$asd-*commands without rewriting paths, URLs, or longer identifiers; Claude-generated skills retain/asd-*commands. - Session recovery finds archived pre-merge sprints without reviving completed or malformed archives.
- Design and implementation review routing, incremental External Review scope, and wrapped-CLI availability reporting now match the canonical cross-provider contracts.