Claude Code · Source-level changelog

Version 1.0.82

Version 1.0.82 introduces real-time validation for Claude Code settings files, simplifies the permission system by removing complex shortcuts, and adds internal infrastructure for upcoming beta features. This is primarily a quality and stability release focused on preventing configuration errors.

Package @anthropic-ai/claude-code Diff v1.0.81 → v1.0.82
4
Features & Changes
0
Bug Fixes
0
In Development
0
Env Vars / Flags

Changes

4 items

Settings File Validation #

What: Real-time validation when editing .claude/settings.json and .claude/settings.local.json files How to use:

# When you edit a settings file with invalid JSON or schema
claude edit .claude/settings.json
# If the edit would create invalid settings, you'll see:
# "Claude Code settings.json validation failed after edit"
# followed by specific errors and the full schema

Details:

  • Validates JSON syntax before applying edits
  • Checks against the official Claude Code settings schema
  • Shows specific field paths and error messages when validation fails
  • Provides the complete schema reference for troubleshooting
  • Prevents invalid configurations from being saved

Simplified Permission System #

What changed: Removed complex permission shortcuts and modes Previous behavior: Had "accept-project" mode and shift+a keyboard shortcut for batch permissions New behavior: Streamlined to three clear options: accept-once, accept-session, or reject Impact: Clearer and more predictable permission handling without confusing shortcuts

Beta Feature Infrastructure #

What changed: Added internal filtering for experimental features Previous behavior: All experimental features handled uniformly New behavior: Special handling for "interleaved-thinking-2025-05-14" and "context-1m-2025-08-07" features Impact: Prepares for controlled rollout of beta features (not yet user-accessible)

Input Context Preservation #

What changed: Enhanced command processing to preserve preceding input blocks Previous behavior: Commands might lose context from previous interactions New behavior: Better preservation of conversation context when executing commands Impact: More consistent behavior when using commands in longer conversations