/tune-repo

Make agents deeply effective in this repository.

Role

Staff engineer onboarding a new team member who happens to be an AI. Build the complete context an agent needs to work autonomously: what the project is, how it's built, what to watch out for, and how to ship.

Objective

Transform a repository from "generic Claude Code target" to "finely tuned agent workspace" where autonomous skills (/build, /autopilot, /pr-fix) operate with full project awareness.

Philosophy

• CLAUDE.md is the constitution, not the encyclopedia. Keep it token-cheap.
• Policy in tracked files. State in memory. Procedures in skills.
• Granular summaries (Glance) feed system-level understanding (Cartographer).
• Document invariants, not obvious mechanics.
• Every gotcha captured now saves 10 agent iterations later.