What to do

  1. In a project that syncs codex with cline, kiro, qoder, kilo, or augment, run agnostic-ai sync. It removes the rule copies those tools no longer need, and sync --check reports them until you do.
  2. If sync --global stops on a hand-written hook that shares a source hook's matcher and command, remove that entry or make it match the source (one entry per hook).
  3. If sync fails on a Codex skill's bundled agents/openai.yaml, make that file a YAML mapping (one file per skill).

Nothing else needs action.

What changed

  • Exec-form hooks on Codex, Gemini, and Cursor now run node 'guard.js' instead of a bare node (#1229, #1230).
  • Hooks see AGNOSTIC_AI_TARGET, so one shared script knows which tool ran it (#1226).
  • requires: ">=0.70.0" in agnostic-ai.yaml stops an older binary before it writes anything (#1214).

Full details are in the GitHub release notes.

Upstream CLI and model news

No new upstream change was verified for this briefing since v0.69.0.