RPI is a meta-orchestrator that runs a complete three-phase lifecycle: discovery, implementation (crank), and validation. It delegates work to dedicated phase skills and enforces strict delegation, complexity classification, execution gates, and automated retry loops so agents run full end-to-end tasks autonomously. Concrete use: run /rpi "<goal>" to convert a high-level objective into a research plan, implement waves of changes, and validate results with post-mortem and retro.
Use RPI when you need a hands-off, full-stack execution of a development or research objective that requires planning, iterative implementation, and validation. Ideal for feature builds, migrations, or cross-cutting refactors where you want repeatable phase gates, automated retries, and reportable artifacts.
Designed for general coding-agent runners (Claude Code, Codex runners, Cursor integrations) and executor environments that support skill delegation and filesystem artifacts.
This skill has not been reviewed by our automated audit pipeline yet.
