Vibeleaderboard
← All Intel
Intel / repo

claude-code v2.1.328

Source
github.com
Author
claude-code-publisher[bot]Top Viber
Date

Context

Claude Code 2.1.328 changes how sessions start for developers who use Claude through a cloud provider instead of directly from Anthropic. Claude Code is Anthropic's coding agent, which runs commands and edits files on a developer's machine and asks first before actions its rules do not already allow. According to the release notes, auto mode is now available for Claude Opus 4.6 and Sonnet 4.6 on Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry and the Claude apps gateway, and an interactive session with no permission mode set now starts in it.

In auto mode, a classifier rather than the person at the keyboard decides whether each action may run. For a developer on one of these providers who never chose a permission mode, the practical effect is fewer approval prompts: actions they used to approve themselves may now be approved by the classifier. The same release makes and Bash permission prompts for these two models on third-party providers use and offer auto mode, which the notes say matches newer models. The `disableAutoMode` setting turns off both of those prompt changes. Because the new default applies only to sessions with no permission mode, choosing a mode explicitly also keeps a session out of it.

A related fix matters to anyone who tries the mode behind a gateway. On Bedrock, Agent Platform, Foundry or the Claude apps gateway, auto mode failed on every turn outside Claude Desktop when `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS` was set and the gateway rejected beta features; 2.1.328 fixes that. The notes do not say what the classifier allows or blocks on these providers, whether it behaves differently from sessions on Anthropic's own API, or whether non-interactive runs such as `claude -p` change their default. They describe the new starting mode only for interactive sessions.

Key takeaways · AI-distilled
  • Code Mode runs faster in auto and bypass-permissions modes: v2.1.328 removes a 0.75-second pause after most scripts, and an edit that changes nothing no longer stops a script.
  • /code-review now reports only bugs and CLAUDE.md violations on most models, marks each finding Important, Pre-existing or Nit, and ends with a suggested fix; reuse and efficiency cleanups move to /simplify.
  • In zsh, Bash commands containing a word that starts with = (such as echo ==== or [ a == b ]) no longer fail with a "not found" error, because =cmd no longer expands to a command's path.
  • Self-hosted runners gain --repository-folder, ---label and --metadata options plus new orchestrator metrics, and no longer stop every session when a proxy answers a poll with HTTP 200 and a body such as {}.
  • Automatic now rebuilds the the way /compact does, instead of reusing the session's earlier system prompt.
Terms in this piece · Glossary
  • plan mode — A read-only agent mode that investigates and proposes a plan without editing anything, so you approve the approach before any code changes.
  • agent hook — A script the harness runs automatically at a fixed point in an agent's loop — before a tool call, after an edit, on session start.
  • context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.
  • system prompt — The standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.
More from claude-code-publisher[bot]
Recommended reads
Comments

Checking sign-in…

Loading comments…