Claude Code v2.1.301 adds retry limits, vim bracket matching and new telemetry fields

The release adds CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS, which lets unattended retry mode end a request when a usage limit resets too far away instead of waiting. OpenTelemetry tool_decision events now include classifier_verdict and permission_mode, so auto mode classifier decisions can be told apart from permission rules. With OTEL_LOG_TOOL_DETAILS=1 set, classifier_category names the built-in rule a classifier block matched.

Several fixes change behavior that agent builders may rely on. The fable and best model aliases now retry once on the previous Fable model when the newest one is not available to an organization. Switching to a model with a smaller context window now has the previous model summarize a conversation that is too long for it first. Quitting no longer removes a worktree made by a WorktreeCreate hook without asking when it may hold unpushed commits.

Grep and Glob now return an error when the search program fails, instead of reporting no matches. The Read tool no longer reads binary files such as .xlsm and .parquet as text. Vim mode gained the % motion for jumping to a matching bracket, and the VS Code extension now renders LaTeX math in responses.

If your agent treats an empty Grep or Glob result as no matches, it should now also handle the error returned when a search fails.

Unattended runs that use the retry watchdog can cap how long a request waits at a usage limit.

Teams reading OpenTelemetry tool_decision events can now separate classifier decisions from permission rule decisions.