[Release] anthropics/claude-code v2.1.143: v2.1.143

Claude Code v2.1.143 introduces plugin dependency enforcement, projected context costs in the marketplace, and improved background session management. Key fixes address credential corruption, PowerShell execution policies, stop hook infinite loops, and Windows Terminal compatibility issues. The release enhances session persistence across idle states and adds new configuration options for agent dispatching.

Key Points

• •Plugin dependency enforcement: `claude plugin disable` now prevents disabling plugins that other enabled plugins depend on, with helpful disable-chain hints; `claude plugin enable` auto-enables transitive dependencies
• •Projected context costs displayed in `/plugin` marketplace browse pane showing per-turn and per-invocation token estimates for cost-aware plugin selection
• •New `worktree.bgIsolation: "none"` setting allows background sessions to edit working copy directly without `EnterWorktree` for repos where worktrees are impractical
• •PowerShell tool now passes `-ExecutionPolicy Bypass` by default on Windows for Bedrock/Vertex/Foundry users; opt out with `CLAUDE_CODE_POWERSHELL_RESPECT_EXECUTION_POLICY=1`
• •Background sessions now preserve model and effort level settings after waking from idle, improving session continuity
• •Fixed corrupt `.credentials.json` with non-array `scopes` value causing CLI startup hangs or silent OAuth token refresh failures
• •Stop hooks now end turns with warning after 8 consecutive blocks (configurable via `CLAUDE_CODE_STOP_HOOK_BLOCK_CAP`) to prevent infinite loops
• •`claude agents` dashboard now accepts `--add-dir`, `--settings`, `--mcp-config`, `--plugin-dir`, `--permission-mode`, `--model`, and `--effort` flags for consistent session configuration
• •Fixed background sessions capturing IDE file references into warm spare input, causing references to prepend to next prompt
• •Improved Windows Terminal compatibility: fixed right-click paste, stale-fragment rendering, and repeated PowerShell process spawning in agent view

Workflow Diagram

Concepts

Artifacts (5)

claude --bg --mcp-config <file> --settings <file> --add-dir <path> --plugin-dir <path> --fallback-model <model> --allow-dangerously-skip-permissions --dangerously-skip-permissions

# Opt out of PowerShell -ExecutionPolicy Bypass
CLAUDE_CODE_POWERSHELL_RESPECT_EXECUTION_POLICY=1

# Disable PowerShell tool on Windows
CLAUDE_CODE_USE_POWERSHELL_TOOL=0

{
  "worktree": {
    "bgIsolation": "none"
  }
}

# Override default 8-block cap for stop hook warnings
CLAUDE_CODE_STOP_HOOK_BLOCK_CAP=15

claude agents --add-dir <path> --settings <file> --mcp-config <file> --plugin-dir <path> --permission-mode <mode> --model <model> --effort <level> --allow-dangerously-skip-permissions