Our Claude Code Cheatsheet (neon.com)

Neon engineer shares a compact Claude Code cheatsheet of hands-on workflows, shortcuts, and add-ons that turned everyday usage into a repeatable developer toolchain. Key practical tips include using Plan Mode (Shift+Tab+Tab) to outline solutions before coding, spinning up parallel task agents (up to ~15) to review code from different lenses (security, style, perf), and keeping large inputs as repo files referenced with @filename or pasted images via Ctrl+V. Essential session commands: /clear, /compact, /context, plus resume (claude -c or claude --resume), /rewind, and a YOLO flag (--dangerous-skip-permissions) for fast sandboxed work. The cheatsheet also highlights memory files (CLAUDE.md by default, AGENTS.md in PRs), custom slash commands in .claude/commands, and a scriptable /statusline for repo/branch/model/token visibility. The post matters because it translates features into reproducible engineering practices that improve code quality and team workflows. Technical implications: using types-first prompts (Zod/DB schema) as guardrails reduces hallucinations and drift; choosing think modes (think → ultrathink) plus token windows (4k–32k) trades latency/cost for deeper cross-file reasoning; headless mode enables CLI-driven automation (e.g., piping npm audit output into Claude for prioritized reports); and plugins like Serena, Context7, and Playwright add semantic refactoring, live docs lookup, and UI testing. Combined, these patterns make Claude Code a scriptable, reviewable, and team-shareable assistant for large refactors, prompt testing, and CI-friendly analysis.