This repository contains a Claude Code configuration based on Boris Cherny's X thread about how he uses Claude Code.

Boris is the creator of Claude Code at Anthropic, and this setup reflects his personal workflow.

Project-specific instructions that Claude reads on startup. Update this file whenever Claude makes a mistake so it learns not to repeat it.

• Pre-allowed permissions: Common safe commands (npm, git, gh, etc.) won't prompt for approval
• PostToolUse hook: Auto-formats code after Write/Edit operations

Based on Boris Cherny's thread (Jan 2026) sharing tips sourced directly from the Claude Code team.

• Run 3-5 Claude sessions in parallel using git worktrees
• Use subagents to throw more compute at problems
• Offload tasks to subagents to keep your main context clean
• Route permission requests to Opus 4.5 via a hook to auto-approve safe ones

• Start complex tasks in plan mode (shift+tab)
• When things go sideways, re-plan instead of pushing through
• Use plan mode for verification steps, not just builds

• Invest in your CLAUDE.md — update it after every mistake
• Create reusable skills and commit them to git
• Use /statusline to show context usage and git branch
• Color-code and name terminal tabs, one per task/worktree

• Challenge Claude: "Grill me on these changes"
• Demand proof: "Prove to me this works"
• Reset mediocre work: "Scrap this, implement the elegant solution"
• Write detailed specs to reduce ambiguity

• Paste Slack bug threads and just say "fix"
• Say "Go fix the failing CI tests" — don't micromanage how
• Point Claude at docker logs to troubleshoot distributed systems
• Use Claude for analytics — works with any database CLI, MCP, or API

• Enable "Explanatory" output style in /config to learn the why
• Have Claude generate visual HTML presentations for unfamiliar code
• Ask Claude to draw ASCII diagrams of protocols and codebases
• Use voice dictation (fn x2 on macOS) — you speak 3x faster than you type

• The team recommends Ghostty for its synchronized rendering and unicode support

Based on Boris Cherny's thread (Mar 2026) sharing his favorite hidden and under-utilized Claude Code features.

• Claude Code has a mobile app (iOS/Android) — open the Claude app and use the Code tab
• Use /teleport or claude --teleport to continue a cloud session on your local machine
• Use /remote-control to control a local session from your phone or browser
• Set "Enable Remote Control for all sessions" in /config for always-on access

• /loop runs a skill on a recurring interval (e.g., /loop 5m /babysit for auto code review and rebase)
• /schedule runs Claude on a cron schedule, up to a week at a time
• Example loops: auto-address code review, auto-rebase PRs, sweep post-merge comments, prune stale PRs
• Turn workflows into skills, then loop them for hands-free automation

• Use hooks to run deterministic logic at each stage of the agent lifecycle
• SessionStart — dynamically load context when Claude starts
• PreToolUse — log every bash command the model runs
• PermissionRequest — route approval prompts to WhatsApp or other channels
• Stop — poke Claude to keep going whenever it stops
• Docs: https://code.claude.com/docs/en/hooks

• Cowork Dispatch — secure remote control for Claude Desktop; catch up on Slack, emails, manage files from mobile
• Chrome extension — connect Claude to your browser for frontend work; Claude iterates until the result looks right
• Desktop app — auto-starts web servers and tests them in a built-in browser

• /branch forks your current session; or use claude --resume <session-id> --fork-session from CLI
• /btw answers quick side questions without interrupting the agent's current work
• /voice enables voice input — hold space bar in CLI, or use the voice button in Desktop

• claude -w starts a new session directly in a git worktree
• /batch fans out massive changesets to dozens, hundreds, or thousands of worktree agents
• Use WorktreeCreate hook for non-git VCS worktree creation

• --bare speeds up SDK startup by up to 10x — skips auto-loading CLAUDE.md, settings, MCPs
• --add-dir (or /add-dir) gives Claude access to additional repos; also grants permissions there
• --agent=<name> runs a custom agent defined in .claude/agents/ — works for non-interactive mode too
• Add "additionalDirectories" to settings.json to always load extra folders

Copy the .claude/ folder and CLAUDE.md to the root of your existing project:

cp -r /path/to/bcherny-claude/.claude /path/to/your-project/
cp /path/to/bcherny-claude/CLAUDE.md /path/to/your-project/

Clone this repo and use it as a starting point for a new project:

git clone <this-repo> my-new-project
cd my-new-project
rm -rf .git
git init

Copy only specific files you want:

• Just want slash commands? Copy .claude/commands/
• Just want subagents? Copy .claude/agents/
• Just want permissions? Copy .claude/settings.json

1. Customize CLAUDE.md for your project's tech stack, conventions, and workflow
2. Edit settings.json to match your project's commands (e.g., bun instead of npm)
3. Run Claude Code in your project directory - it will automatically read the config

In Claude Code, type / to see available commands:

/commit-push-pr     # Full git workflow
/quick-commit       # Fast commit
/test-and-fix       # Run and fix tests
/review-changes     # Code review
/worktree           # Parallel Claude sessions
/grill              # Adversarial code review
/techdebt           # Codebase cleanup

Ask Claude to use a subagent:

"Use the code-simplifier agent to clean up the code I just wrote"
"Run the verify-app agent to test everything works"
"Use code-architect to review this design"

• Original setup: https://x.com/bcherny/status/2007179832300581177
• Team tips (Jan 2026): https://x.com/bcherny/status/2017742741636321619
• Power features (Mar 2026): https://x.com/bcherny/status/2038454336355999749