Claude Code v2.1.101: /team-onboarding

What Changed — The Specific Update

Claude Code v2.1.101 dropped with over 40 changes, but three stand out for daily users: a new /team-onboarding command, major fixes to --resume session recovery, and default OS CA certificate store trust for enterprise environments.

The /team-onboarding command analyzes your local Claude Code usage patterns and generates a customized ramp-up guide for new teammates. It's like having Claude write your team's internal documentation based on how you actually work.

Session resumption fixes address critical bugs where --resume and --continue would lose context on large sessions, bridge into unrelated subagent conversations, or crash when persisted tool results were malformed. These were pain points for developers working on complex, multi-session projects.

Enterprise TLS proxy support now works out-of-the-box by default trusting the OS certificate store. Previously, you'd need manual configuration; now it just works. If you need the old behavior, set CLAUDE_CODE_CERT_STORE=bundled.

What It Means For Your Daily Workflow

1. Onboard Teammates Faster

Run /team-onboarding in any Claude Code session. Claude will analyze your project structure, common commands, and workflow patterns, then generate a tailored guide. This follows Anthropic's recent focus on team collaboration tools, as seen in their April 4th integration framework announcement connecting to tools like Asana and Slack.

2. Resume Sessions Without Losing Context

The --resume fixes mean you can now reliably pick up where you left off, even with:

• Large sessions that previously anchored on dead-end branches
• Subagent conversations that would previously cause chain recovery issues
• Sessions created with -p or the SDK (fixed claude --continue -p)

Use claude -p --resume <name> with session titles set via /rename or --name for better organization.

3. Improved Focus and Brief Modes

Focus mode now generates more self-contained summaries since Claude knows you only see the final message. Brief mode retries once when Claude responds with plain text instead of structured output. These subtle improvements make single-purpose interactions more efficient.

4. Better Error Messages and Tool Control

• Tool-not-available errors now explain why a tool exists but isn't available in the current context
• Rate-limit messages show which limit was hit and when it resets, not just a countdown
• Permission rules now correctly override hooks: permissions.deny beats a PreToolUse hook's permissionDecision: "ask"
• MCP tools properly inherit to subagents from dynamically-injected servers

Try These Commands Now

# Generate a team onboarding guide from your current project
/team-onboarding

# Resume a session by its custom name (after using /rename)
claude -p --resume "auth-refactor-session"

# Force bundled certificates if OS trust causes issues
export CLAUDE_CODE_CERT_STORE=bundled
claude code

# Check for plugin updates with better warnings
/plugin update

# Use improved brief mode for concise responses
/brief

Other Notable Fixes

• Memory leak fixed: Long sessions no longer retain dozens of historical message list copies
• Hardcoded timeout removed: Slow backends (local LLMs, extended thinking) respect API_TIMEOUT_MS
• Grep tool self-heals: Falls back to system rg if embedded binary path becomes stale
• Remote Control improvements: Better SSH handling and crash recovery
• Plugin system fixes: Directory-source plugins load fresh, skills honor context fields

This release follows Claude Code's April 8th introduction of Tool Search for MCP token savings and aligns with their ongoing focus on stability and enterprise readiness.