Onboarding Cc

Instructions

Use this skill to make Claude Code productive in a large, long-lived repository by turning “context” into maintained project artifacts.

1. Create a durable context home

• Maintain a dedicated, versioned place for AI context that can improve over time (the post’s example is a separate repository).
• Keep it stable across branch churn and long-running release lines.

2. Add a repo-root onboarding entrypoint

• Add a CLAUDE.md at the repository root.
• Include:
  • Environment setup steps
  • The “lay of the land” (where key code lives)
  • Links to deeper documentation (“reference, don’t embed”)

3. Build a small skill library for recurring work patterns

• Create reusable skills that encode your team’s expectations.
• Keep skills short and point to canonical docs rather than duplicating them.
• Examples mentioned in the post include skills for:
  • Project orientation
  • Version control conventions
  • Debugging (load when investigating bugs/failures)

4. Use integrations to reduce manual context-passing

• Where appropriate, add integrations that can expose project signals to Claude Code (e.g., tests, exceptions, support threads, release tags).
• Prefer automation that produces summaries and diffs over raw dumps.

Examples

Example: Minimal onboarding checklist for a new repo

• Create or update CLAUDE.md with environment setup and links to docs.
• Add/update small skills for debugging and version control.
• Add integrations for the project’s “health signals” (tests, exceptions, support).

Example: “Reference, don’t embed” documentation pattern

Instead of copying a long internal document into an instruction, add a short pointer:

• “See docs/architecture.md for module boundaries.”
• “See docs/testing.md for how to run and interpret nightly tests.”

Source

• https://claude.com/blog/onboarding-claude-code-like-a-new-developer-lessons-from-17-years-of-development