You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
ElsewhereArea: what it touches
ImprovementsKind: in v2.1.299,
ImprovementsSection of the release
What
--safe-mode is a flag (an option added when starting Claude Code) that starts it with your customizations switched off, which helps when a broken setting is causing trouble. Several things about it changed:
Agent list footer: the footer of the agents list now shows "safe mode" and "customizations are off" next to the waiting line. Before, only the waiting text appeared.
Restarts: --safe-mode is now one of the flags carried over when a background job or session restarts. The warning about flags dropped on restart now reads "stripped respawnFlags token(s) that a restart does not re-apply" instead of calling them "non-allowlisted", and the saved flags go through an extra clean-up step.
Exit hint: the hint telling you how to leave safe mode is now produced on the fly instead of always being the fixed text "restart without --safe-mode". Claude Code checks whether --safe-mode was on the command line, and the evidence includes the text "unset CLAUDE_CODE_SAFE_MODE", so the hint can point to the CLAUDE_CODE_SAFE_MODE environment variable as well. The matching hint for bare mode, which used to be the fixed text "restart without --bare", is built the same way.
Why
You can now see at a glance that a session is in safe mode, a background job started in safe mode stays in safe mode after it restarts, and the exit hint can reflect how safe mode was actually turned on.
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed sinceAnthropic's documentation has since written up --safe-mode, on What's new.Also this week: **sub-agents can spawn their own sub-agents** (background chains are capped at five levels deep); **`--safe-mode`** starts Claude Code with all customizations disabled for troubleshooting; and **`fallbackModel`** configures…whats-new/indexsee the edit
How sure we are
Two sources agreeTwo things we can check say the same as this entry.
Anthropic's documentation agreesAnthropic's documentation has since written up --safe-mode, on What's new.
Anthropic's release notes agreeAdded a "safe mode" line to the agents view peek panel for background sessions started with --safe-mode
