Claude Code v1.0.60

Agent Management System (New!)

• New /agents command: Create and manage custom subagents that Claude can delegate tasks to
• Create agents with specialized system prompts and tool access
• Choose between AI-generated or manual agent configuration
• Agents can be stored at project level (.claude/agents/) or personal level (~/.claude/agents/)
• Built-in agents are always available and cannot be modified
• Each agent can have a custom background color for visual distinction

Example usage:

# Open the agent management interface
/agents

Background Shell Management Tools (New!)

• KillShell Tool: Kill a running background shell by its ID
• Use /bashes command to find shell IDs
• Returns success/failure status

• BashOutput Tool: Retrieve output from background shells
• Returns only new output since last check
• Shows stdout, stderr, and shell status
• Automatically truncates very long outputs

Bash Output Summarization

• Claude now intelligently summarizes verbose command outputs
• Summaries are triggered for outputs over 5000 characters
• Original output is preserved and can be viewed in verbose mode
• Includes sections for: Overview, Detailed Summary, Errors, and Verbatim Output
• Particularly useful for build logs, test results, and repetitive debug output

Example scenario:

# Running a command with verbose output
npm run build

# Claude will automatically summarize if output is lengthy, showing:
[Summarized output]
Overview: Build completed successfully...
Detailed summary: ...
Errors: None encountered
Verbatim output: [key snippets preserved]

Spinner Tips

• Tips functionality has been made configurable
• Tips are now fetched dynamically based on user preferences
• Improved tip management system

Enhanced Dialog Components

• New bv1 component for better dialog presentation
• Improved escape key handling with "Press Esc again to exit" confirmation
• Consistent styling with rounded borders and proper spacing

OAuth Flow Improvements

• Added manual code entry support for OAuth authentication
• Better error handling with retry options
• Visual improvements during the authentication process
• Support for long-lived tokens (1 year expiry) for GitHub Actions

️ Command System Enhancements

• New agent type system with built-in types:
• general-purpose: General multi-step task handling
• context-collector: Codebase exploration and metrics
• tech-researcher: Technical solution research
• test-engineer: Test writing and maintenance
• spec-designer: Specification refinement
• file-organizer: File operations and organization
• comment-writer: Code documentation
• investigator: Deep code investigation
• refactoring-specialist: Code improvement
• code-auditor: Architecture and security analysis
• implementation-engineer: Feature implementation
• documentation-specialist: Comprehensive documentation
• style-fixer: Code formatting and style

Aliases and Usability

• /clear command now also responds to /reset alias
• Improved permission handling for additional working directories
• Better path resolution and validation

Agent Creation Workflow

1. Choose location: Project or personal agents directory
2. Creation method: AI-generated or manual configuration
3. Define purpose: Describe what the agent should do
4. Select tools: Choose which tools the agent can access
5. Customize appearance: Optional background color
6. Review and save: Confirm configuration before saving

Example agent creation:

/agents
# Select "Create new agent"
# Choose location: Project
# Method: Generate with Claude
# Describe: "Expert code reviewer that checks for security vulnerabilities and best practices"
# Claude generates the configuration
# Select tools the agent should have access to
# Choose a background color
# Save the agent

• Removed unused OAuth error classes and related error handling code
• Cleaned up shell snapshot cleanup functionality
• Removed various deprecated functions and imports
• Fixed issue with pasted text display (now shows "[Pasted text #X]" without "+0 lines")

• None identified in this release

• Agents provide a powerful way to create specialized assistants for specific tasks
• Background shell management allows better handling of long-running processes
• Output summarization helps maintain context efficiency while preserving important information
• The new agent system is particularly useful for repetitive tasks like code reviews, testing, and documentation