Guard Aurora action listing against recursion

[brunda-coder]

Motivation

• Prevent an application crash caused by a RecursionError raised while deriving Aurora canvas state when enumerating actions via list_actions() in compute_canvas_state().

Description

• Add a defensive fetch in _filter_actions (core/aurora_state.py) that attempts registry.list_actions() and falls back to getattr(registry, "_actions", []) on RecursionError or other exceptions, preserving the existing filtering, sorting and conversion to ActionItem.

Testing

• No automated tests were executed for this small runtime guard change.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2970d15357

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Emit ACTION_FAILED when dispatch returns ok=False

Because handle_intent always emits ACTION_FINISHED after _dispatch_action, failures (e.g., workspace_save_snapshot when no workspace exists or os_open_workspace_folder without an OS pipeline) will still be reported as successful actions. Downstream consumers like AvatarBehaviorEngine treat ACTION_FINISHED as a success signal, so the UI/voice feedback will incorrectly celebrate failed actions. Consider emitting ACTION_FAILED (and/or skipping ACTION_FINISHED) when result.get("ok") is false to keep event semantics consistent with actual outcomes.

Useful? React with 👍 / 👎.