Codex CLI: Cursor Skill Import & Duplicate-Free Session Sync
CodexCodex CLI 0.147.0 expanded the /import command so it can now pull in Cursor-managed skills, not just the settings, MCP servers, plugins, and sessions it already migrated. The release also adds duplicate-free sync, so re-running /import against a previously imported Claude Code or Cursor conversation updates it in place instead of creating a repeat copy. Together the changes make it more reliable for developers who run Codex alongside Cursor or Claude Code to keep their setups in sync without manual cleanup.
Key Takeaways
- Cursor skills are now portable into Codex, closing a gap where only MCP servers, plugins, sessions, and memories could previously be imported from Cursor.
- Duplicate-free sync means re-running
/importon a previously migrated Claude Code or Cursor conversation now updates it in place instead of creating a repeat thread. - The change builds directly on the 0.145.0 Cursor import support shipped in July 2026, rather than introducing a new migration path from scratch.
- It targets developers running a mixed toolchain, where Cursor, Claude Code, and Codex CLI are all in active use side by side.
- The feature shipped bundled with 0.147.0's thread-organization and
--approve-for-mework, both of which are already tracked separately in the changelog collection. - Third-party coverage is thin — as of this run only the official GitHub release notes and Releasebot's aggregator page cover the feature, with no Hacker News, Reddit, or YouTube discussion found yet.
Sources & Mentions
2 external resources covering this update
Cursor Skills Join the Import Pipeline
Codex CLI's /import command has been steadily expanding since it first shipped Claude Code migration support earlier in 2026, later adding Cursor configuration support in the 0.145.0 release. With Codex CLI 0.147.0, the command takes another step forward: it can now import Cursor-managed skills directly, alongside the MCP servers, plugins, sessions, slash commands, and project-scoped memories it already handled.
This closes a gap for developers who have built up a library of reusable skills inside Cursor and want to bring that investment with them rather than recreating it by hand inside Codex. Skills are one of the more time-consuming assets to rebuild manually, so making them portable lowers the switching cost for teams evaluating Codex against other AI coding agents.
No More Duplicate Imports
Alongside the expanded skills support, Codex CLI 0.147.0 changes how repeated imports behave. Previously, re-running /import against a Claude Code or Cursor session that had already been imported could create a duplicate entry in Codex's thread list. Now, Codex synchronizes changes to already-imported conversations instead of duplicating them, so developers can safely re-run /import as their source tool's history changes without cluttering their Codex workspace.
Why It Matters
Together, these two changes make /import behave less like a one-time migration tool and more like an ongoing sync mechanism between Codex and the other AI coding tools a developer may still be using day to day. For teams running a mixed toolchain — Cursor for some work, Claude Code or Codex for other work — this reduces the friction and manual bookkeeping involved in keeping context consistent across tools.
The feature shipped in the same 0.147.0 release as expanded thread organization (persistent sections, incremental transcript loading) and the new --approve-for-me flag for automated approval reviews, continuing Codex CLI's broader push toward smoother, less interruption-prone long-running sessions.