Source Intelligence

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 olderv2.0.11 v2.0.13newer

Claude Code v2.0.12

7 entries read diff v2.0.11 → v2.0.12 Markdown

Version 2.0.12 is a maintenance release focused on internal code reorganization and plugin system enhancements. The primary changes involve plugin command configuration improvements and minor UI refinements. Most of the 12,384-line increase is due to bundler regeneration with different variable names (~6,898 renames), not new functionality.

Find
Pick an entry · j / k steps through
7 entries

Changesopen

Mode Commands System#

What: Commands can now be designated as "mode commands" for specialized, multi-phase workflows

How to use: Add mode: true to command frontmatter in your .claude/commands/*.md files:

mode: true
description: Guided workflow for complex refactoring

Your command content here...

Details:

  • Mode commands receive special priority in Claude's tool selection
  • Display distinct permission UI titled "Use guided workflow" instead of "Slash command"
  • Show message "Your request appears to match a specialized workflow"
  • Intended for complex, multi-step tasks that benefit from structured guidance
  • Evidence: isModeCommand: K at line 440835 in loadCommands() function, mode command filtering in V6Q() function

Inline Plugin Commands#

What: Plugin developers can now define commands directly in their manifest with inline markdown content

How to use: In your plugin's marketplace-manifest.json:

{
  "commands": {
    "hello": {
      "content": "---\ndescription: Say hello\n---\n\nGreet the user warmly!",
      "description": "Greet the user",
      "argumentHint": "[name]"
    },
    "status": {
      "source": "commands/status.md"
    }
  }
}

Details:

  • Commands can now use object notation with content (inline) or source (file path) fields
  • Cannot use both content and source - validation enforces one or the other
  • Supports metadata overrides: description, argumentHint, model, allowedTools
  • Reduces file clutter for simple commands
  • Old array-of-paths format still supported for backward compatibility
  • Evidence: Command schema with content field at lines 42750-42775 in schema definition, inline content processing in plugin loader at lines 48817-48850

Enhanced Plugin Command Validation#

Better error handling when plugin manifests contain invalid command entries. The system now:

  • Validates that array-format commands contain only strings
  • Logs descriptive errors for malformed entries: "Unexpected command format in marketplace entry for {plugin-name}"
  • Skips invalid entries instead of crashing
  • Evidence: Type validation in plugin loader at lines 72627-72633 in loadPluginsFromMarketplace() function

Cleaner Slash Command Permission UI#

The permission confirmation prompt for slash commands now displays more user-friendly information:

  • Title changed from generic "Tool use" to specific "Slash command"
  • Shows the actual command string instead of SlashCommand(command-name)
  • Displays command description instead of technical Execute slash command: /command
  • Includes command metadata for better context-aware routing
  • Evidence: UI component changes at lines 72425-72430 in permission confirmation renderer, metadata addition at line 65680

Internal Code Organization#

Significant internal restructuring for improved maintainability:

  • Ripgrep module relocated from line ~335k to line ~3.8k for faster loading in --ripgrep mode
  • Entry point handling simplified with better separation of concerns
  • Changed from process.exit() to process.exitCode = for more graceful shutdown
  • ~6,898 variable renames throughout codebase (no functional impact)
  • Evidence: Entry point refactoring at lines 461763-461773 in az8() function, ripgrep module relocation to lines 3856-3869

Related

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

Bundle Changes#

  • File size increased from 449,389 to 461,773 lines (+2.7%)
  • Structural similarity: 75.0% (25% appears different due to variable renaming)
  • Declaration changes: 1,207 additions, 2,304 deletions, 102 modifications, 6,898 renames
  • No dependency version upgrades - all bundled libraries remain at same versions
  • Bundler regeneration caused most of the diff (not actual code changes)

Related

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

What This Release Does NOT Include#

  • No new command-line flags or options
  • No changes to debug flags (--debug, -d, --debug-to-stderr, -d2e remain identical)
  • No ripgrep configuration changes (system vs builtin selection unchanged)
  • No dependency upgrades (Axios, Zod, Fuse.js, LRU-cache, Minimatch, Lodash all unchanged)
  • No new error handling or validation for end users
  • No changes to core CLI functionality

Related

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

1 entry

Notes for Plugin Developersopen

#

If you maintain Claude Code plugins, you can now:

  1. Use mode: true frontmatter to designate workflow commands
  2. Define simple commands inline in your manifest instead of creating separate .md files
  3. Override command metadata (description, model, allowed tools) at the manifest level
  4. Expect better validation errors if your manifest format is incorrect

The old plugin manifest formats remain fully supported - these are additions, not breaking changes.

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.0.12. 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.

  • Plugin System Released: Extend Claude Code with custom commands, agents, hooks, and MCP servers from marketplaces
  • /plugin install, /plugin enable/disable, /plugin marketplace commands for plugin management
  • Repository-level plugin configuration via extraKnownMarketplaces for team collaboration
  • /plugin validate command for validating plugin structure and configuration
  • Plugin announcement blog post at https://www.anthropic.com/news/claude-code-plugins
  • Plugin documentation available at https://code.claude.com/docs/en/plugins
  • Comprehensive error messages and diagnostics via /doctor command
  • Avoid flickering in /model selector
  • Improvements to /help
  • Avoid mentioning hooks in /resume summaries
  • Changes to the "verbose" setting in /config now persist across sessions
System prompt

The system prompt was not captured for this release, so this page cannot say whether it moved.