Boris Cherny's Claude Code Tips Are Now a Skill. Here Is What the Complete Collection Reveals.

The Technique

Boris Cherny, a prominent developer and early Claude Code adopter, has compiled his hard-won tips for effective agentic coding into a single, shareable resource called a Claude Skill. This isn't just a blog post; it's a structured collection of prompts, CLAUDE.md configurations, and workflow patterns that you can import directly into your Claude Code environment. The Skill encapsulates strategies for everything from initial project setup to complex refactoring tasks.

Why It Works

Cherny's tips work because they are battle-tested heuristics that address common friction points when working with AI coding agents. They provide Claude with clearer context, better guardrails, and more explicit instructions, which reduces back-and-forth and increases the likelihood of correct, usable output on the first try. By formalizing these as a Skill, the knowledge becomes actionable—not just theoretical advice you have to remember and type out manually.

How To Apply It

You can access and use this Skill collection to immediately upgrade your Claude Code setup. Here’s how:

1. Locate the Skill: The collection is available through Claude's Skill discovery system. Search for "Boris Cherny" or "Claude Code Tips" within your Claude interface (likely in Claude Desktop or the web app's Skills section).

2. Enable the Skill: Once found, simply enable it. This will make its predefined instructions and patterns available to Claude during your coding sessions, without you having to copy-paste prompts from a document.

3. Key Takeaways to Implement Manually: If you prefer to integrate the concepts directly into your workflow, focus on these core principles from the collection:

   • Structured CLAUDE.md: Go beyond basic instructions. Use sections for ## CONTEXT, ## TECH STACK & RULES, ## WORKFLOW, and ## OUTPUT FORMAT to give Claude a comprehensive playbook for your project.
   • The "Chain of Thought" Command: Prefix complex requests with /think or instruct Claude to "reason step-by-step" before writing code. This leverages Claude's strength in structured reasoning and often yields more accurate implementations.
   • Iterative Scope Control: Break large features into discrete, verifiable subtasks. Instead of "add user authentication," sequence requests like "1. Set up auth.js utility with signIn function," then "2. Create the /api/auth/[...nextauth].js route handler."
   • Explicit Validation Rules: In your CLAUDE.md, specify how code should be validated. For example: "After any change to a React component, run npm test -- ComponentName and report the result." This builds self-correction into the agent's workflow.