1min.

1MIN · AI & AGENTS

Claude Code 2.1.292 fixes permission boundaries and scheduled work

Claude Code 2.1.292 fixes several cases where permission checks, managed settings or background-task state could behave incorrectly. These include network-path reads bypassing a prompt and saved schedules failing after session changes. The release also adds explicit effort settings for subagents. It follows yesterday’s task-recovery update with additional changes to agent control and reliability.

Anthropic · Oct 6, 2026

Why it matters

Existing users who rely on permission policies or recurring tasks have concrete reasons to review this update.

More context

Update through your normal supported installation path and verify a representative permission denial and scheduled task. Plugin installation now waits for managed settings on first use. Access still requires the usual Claude Code account or provider setup; this is a client update.

Keep in mind

These are maintainer-reported fixes, not evidence that unattended agents are now error-free. We have not independently reproduced the bugs or tested the patched behavior.

Original read

More engineering ideas, one minute at a time.

Explore more stories