Source Intelligence
Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All releases Home ←olderv2.1.182 v2.1.185newer→

Claude Code v2.1.183

2 entries read diff v2.1.182 → v2.1.183 Markdown

This release removes the /migrate command and claude migrate CLI that imported configurations from OpenAI Codex and Google Gemini CLI. In its place, the /init command receives a substantial overhaul: it now runs a structured, multi-phase guided interview that sets up not just CLAUDE.md but also personal CLAUDE.local.md files, skills, and hooks in a single session.

Find
Pick an entry · j / k steps through
1 entry

Removed Featuresopen

/migrate command and claude migrate CLI removed#

The /migrate slash command and the claude migrate terminal subcommand have been removed entirely from this version.

Previously these allowed importing MCP servers, slash commands, subagents, skills, and instruction files from:

  • OpenAI Codex (~/.codex/config.toml, AGENTS.md, prompts/)
  • Google Gemini CLI (~/.gemini/settings.json, GEMINI.md, commands/)

The interactive checkbox picker UI, the --dry-run preview mode, the --yes headless mode, and the automatic fallback skill generator are all gone.

Users who previously relied on /migrate to bring over Codex or Gemini settings will need to perform that migration manually by consulting the relevant config files directly.

Evidence

CLI argument parser entry removed (search for "Which agent to import from (codex, gemini)"); /migrate slash command registration removed; dozens of supporting functions deleted including the Codex scanner, Gemini scanner, MCP import logic, subagent writer, and interactive picker component.

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

1 entry

Improvementsopen

/init command — major guided-setup overhaul#

The /init slash command has been significantly expanded. It now runs a structured 8-phase guided interview that covers far more than creating a CLAUDE.md.

Usage
/init

The new flow works as follows:

Phase 0: Checks whether a CLAUDE.md already exists and branches accordingly — offering to review and improve it, leave it alone, or start fresh.

Phase 1: Asks what to set up, with four options:

  • Project CLAUDE.md (team-shared, committed to source control)
  • Personal CLAUDE.local.md (private, gitignored)
  • Both project and personal files
  • "Let Claude decide" — fastest path, no follow-on questions

A second question asks whether to also set up skills and/or hooks.

Phase 2: Launches a subagent to survey the codebase — manifest files, README, CI config, existing AI tool configs (.cursor/rules, .github/copilot-instructions.md, AGENTS.md, .devin/rules/, .windsurf/rules/, .clinerules), formatter config, and git worktree layout.

Phase 3: Asks targeted follow-up questions about anything the code scan couldn't determine — non-standard commands, branch conventions, required env vars, testing quirks, or personal workflow preferences. Synthesizes a proposal listing exactly which files, skills, and hooks will be created, and asks for approval before proceeding.

Phase 4: Writes the project CLAUDE.md using a strict "would removing this cause Claude to make mistakes?" test to keep it concise. Handles the "review and improve" path with diff proposals rather than silent overwrites. For large projects, suggests organizing into .claude/rules/ subdirectory files.

Phase 5: Writes CLAUDE.local.md for personal preferences (role, sandbox URLs, communication style). For repos using sibling/external git worktrees, it writes the content to ~/.claude/<project-name>-instructions.md and creates a one-line stub import in each worktree's CLAUDE.local.md.

Phase 6: Creates skills from the proposal — both skills the user requested during the interview and additional suggestions that fit the codebase (reference knowledge, repeatable deploy or release workflows, etc.). Places each at .claude/skills/<name>/SKILL.md. Distinguishes between skills that can be model-invoked vs. user-only skills with side effects (disable-model-invocation: true).

Phase 7: Checks for GitHub CLI (which gh) and offers to explain its benefits if missing; checks for a lint config and offers to set one up if absent; wires up any requested hooks by calling the update-config skill for proper schema validation and live testing.

Phase 8: Recaps all files written and suggests next steps, including relevant official plugins (frontend-design, playwright, skill-creator) based on what was found in the codebase.

Details
  • The "Review and improve" path in Phase 0 skips the interview and goes straight to proposing specific diffs against the existing file.
  • The "Leave it" path in Phase 0 skips CLAUDE.md entirely and jumps to skills and hooks setup.
  • Q2 (skills/hooks) is a hint rather than a hard filter — if no hook-shaped work exists, /init says so and proposes better-fitting artifacts instead.
  • Hook target file defaults: project choice → .claude/settings.json; personal choice → .claude/settings.local.json. Only asks if the user picked "both".
  • Git pre-commit hooks are correctly distinguished from Claude Code PostToolUse hooks — if a user asks to "gate before committing", /init routes that to a git hook or husky rather than silently creating a mismatched Claude Code hook.
Evidence

New qKp string literal contains the full 8-phase guided setup instructions (search for "Phase 0: Check for an existing CLAUDE.md"). New WKp string literal contains the standalone CLAUDE.md creation instructions for non-interactive contexts (search for "Please analyze this codebase and create a CLAUDE.md file").

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.1.183. Text is unmodified from the upstream changelog. Everything else on this page came out of the bundle instead, which is why the two lists don't match.

  • Improved auto mode safety: destructive git commands (git reset --hard, git checkout -- ., git clean -fd, git stash drop) are now blocked when you didn't ask to discard local work, git commit --amend is blocked when the commit wasn't made by the agent this session, and terraform destroy/pulumi destroy/cdk destroy are blocked unless you asked for the specific stack
  • Added a warning when the requested model is deprecated or automatically updated to a newer model, shown on stderr in print mode (-p) and now also covering models set in agent frontmatter
  • Added attribution.sessionUrl setting to omit the claude.ai session link from commits and PRs in web and Remote Control sessions
  • Added /config --help to list all available shorthand keys for /config key=value
  • Changed /config toggle behavior: Enter and Space both change the selected setting, and Esc now saves and closes instead of reverting
  • Removed the startup "setup issues" line under the logo — run /doctor to see configuration issues or use --debug
  • Fixed thinking.disabled.display: Extra inputs are not permitted 400 errors on subagent spawns and session-title generation for affected configurations
  • Fixed WebSearch returning empty results in subagents
  • Fixed the terminal cursor being stranded above the prompt after navigating history in vim mode with the native cursor enabled
  • Fixed fullscreen TUI corruption (statusline mid-screen, duplicated spinner rows, merged text) in Windows Terminal under heavy nested-subagent load
  • Fixed turns silently completing with no visible output when the model returned only a thinking block; Claude now re-prompts once
  • Fixed user-level skills appearing multiple times in slash-command autocomplete when multiple plugins are enabled
  • Fixed MCP servers requiring authentication exposing auth-stub tools to the model in headless/SDK mode
  • Fixed tmux teammate panes failing to launch when the shell has slow rc-file initialization, and keystrokes typed during agent spawn leaking into the new tmux pane instead of the leader prompt
  • Fixed background tasks started by a teammate being killed when the teammate finishes a turn
  • Fixed scheduled task and webhook trigger deliveries being treated as keyboard input; they now classify as task notifications and can no longer approve a pending action or set the session title in auto mode
  • Fixed focus mode showing "Ran N PostToolUse hooks" timing lines under each response
Read it on GitHub →
System prompt

No change to the system prompt since v2.1.182.

Claude Code, interactive mode

Read the prompts for this release →