Claude Code 2.1.287 → 2.1.296: the changes worth knowing

Claude Code shipped ten releases between 1 and 9 October. The full changelogs are long, so here are the changes most people will notice, each linked to its GitHub release.

New features

• Claude Mods (2.1.287): plugins can now change deeper behaviour. A built-in mod called "You should know" uses a side agent to flag things you or Claude might miss. Enable it with /plugin enable cc-plugin-you-should-know@builtin.
• Get a cleared prompt back (2.1.288): if you clear a prompt with Ctrl+C, press Up on the empty prompt to restore it, pasted text and images included.
• /code-review --max-findings (2.1.288): ask for more or fewer findings than the default, or all of them.
• Managed Agents onboarding (2.1.290): /claude-api managed-agents-onboard <url> turns a page describing a Managed Agents pattern into ant apply files.
• Effort for sub-agents (2.1.292): the Agent tool takes an effort parameter, so a sub-agent can run at the effort level you ask for.
• Haiku 5.5 (2.1.293): claude-haiku-5-5 is now the default Haiku model on the Anthropic API.
• Program Status Protocol (2.1.295): terminals that support OSC 7501 can show whether Claude Code is working, waiting on you or done.
• Bigger reads (2.1.296): the Read tool gets an allow_large option to read a file past the usual size limit in one call, when the context has room.

Fixes that matter

• Mid-response API timeouts no longer fail the turn; non-interactive sessions and subagents continue from the partial response (2.1.288).
• Long sessions with hundreds of images no longer get stuck on "Request rejected as unprocessable" errors (2.1.290).
• Security: PreToolUse hook approvals and auto mode no longer bypass the permission prompt for file reads from network (UNC) paths (2.1.292).
• Claude no longer sometimes treats its own last actions before a compaction as done after it, then retracts or redoes finished work (2.1.293).
• prompt and agent hooks written as instructions, such as "Block commands that...", no longer let through what they should block (2.1.294).
• New onFailure: "block" for command and HTTP hooks: a hook that can't start, times out or exits unexpectedly now blocks the action instead of letting it through (2.1.295).