A structured approach to working with Claude on code, complete with role management and quality gates. You pick driver mode if you're writing and want real-time guidance, navigator mode if you want Claude to implement while you direct, or switch mode to alternate every few minutes. The TDD and review modes enforce test-first workflows and continuous scanning. What's interesting here is the truth-score verification system that can roll back changes if quality drops, plus session persistence so you can pause and resume. Works best when you've got a testing framework and linter already configured. It's more process-heavy than just chatting with Claude, but that's the point: it adds structure to longer coding sessions where roles and quality tracking actually matter.
npx -y skills add spencermarx/open-code-review --skill "Pair Programming" --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
wshobson/agents
dbt-labs/dbt-agent-skills
github/awesome-copilot