Workflow Orchestration

1. Plan Mode Default

• Enter plan mode for ANY non-trivial task (3+ steps or architectural decisions)
• If something goes sideways, STOP and re-plan immediately
• Iterate on the plan until it’s solid, then switch to auto-accept mode
• “A good plan is really important” — Boris Cherny

2. Verification First (Most Important)

• Give Claude a way to verify its work — this 2-3x the quality of results
• Run tests, check logs, open browser, use simulators
• Never mark a task complete without proving it works
• Build domain-specific verification into your workflow

3. Self-Improvement Loop (Compounding Engineering)

• After ANY correction: update CLAUDE.md with the pattern
• Tag @.claude in PRs to auto-update rules
• Review lessons at session start
• The goal: Claude gets better over time, not just per-session

4. Subagent Strategy

• Use subagents to keep main context window clean
• One task per subagent for focused execution
• Common patterns: code-simplifier, verify-app, build-validator

5. Slash Commands for Inner Loops

• Create commands for workflows you do many times a day
• Store in .claude/commands/, check into git
• Embed bash to pre-compute info (git status, etc.)
• Example: /commit-push-pr

6. Hooks for Automation

• Use PostToolUse hooks to auto-format code
• Catches the 10% of formatting issues that slip through
• Prevents CI failures from style issues

7. Permission Management

• Don’t use --dangerously-skip-permissions by default
• Pre-allow known safe commands via /permissions
• Store in .claude/settings.json, share with team

8. Autonomous Bug Fixing

• When given a bug report: just fix it
• Point at logs, errors, failing tests — then resolve them
• Zero context switching required from the user

Core Principles

• Simplicity First: Make every change as simple as possible
• Verify Everything: Feedback loops are more valuable than speed
• Compound Learning: Every mistake becomes a rule