rpi-compound

Orchestrates iterative research, plan, implement, and review loops for software delivery.

5|Updated Sep 24, 2017
One-click install
npx skills add https://github.com/madyankin/dotfiles --skill rpi-compound-madyankin
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rpi-compound
Source: https://github.com/madyankin/dotfiles/tree/main/.config/agents/skills/rpi-compound
Command: npx skills add https://github.com/madyankin/dotfiles --skill rpi-compound-madyankin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? One-shot execution often ships incomplete or risky work. This Skill runs repeated delivery loops that improve correctness, confidence, and execution readiness until explicit completion criteria are met. ## Core Features & Use Cases - Compounding Loop Orchestration: Coordinates rpi-research, rpi-plan, rpi-iterate, rpi-implement, and rpi-review stages with one owner per stage to avoid duplicated work. - Bounded Increments: Executes one phase or tightly scoped slice per loop, keeping automated and manual verification separate. - Compounding Log: Records each loop's goal, work completed, verification results, review outcome, and next action in _scratchpad/plans/compounding-log.md. - Use Case: When implementing a multi-phase feature, run this Skill to plan, implement one phase, review findings classified as critical/major/minor, and iterate until no critical findings remain. ## Quick Start Use the rpi-compound skill to run iterative delivery loops on my current plan until all phases are complete and verified.

Frequently Asked Questions about rpi-compound

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run iterative implementation loops instead of one-shot execution?▼

Use the rpi-compound loop model: research gaps, create or update a plan, implement one bounded slice, then run a review gate. Repeat until plan phases are complete, verification passes, and no critical findings remain.

What is the difference between rpi-plan and rpi-iterate?▼

rpi-plan creates a new executable plan when none exists, while rpi-iterate updates an existing plan that has gone stale. The loop requires a current plan before any implementation step begins.

When should I use rpi-compound instead of running rpi-implement directly?▼

Use rpi-compound when the work spans multiple phases or carries unknowns that benefit from repeated review cycles. Run rpi-implement directly only for a single well-defined slice with an up-to-date plan.

How are review findings classified in each loop?▼

The review gate classifies findings as critical, major, or minor. The classification drives the next action: implement the next phase, iterate the plan, or run targeted research.

When does the compounding loop stop?▼

The loop stops when all plan phases are complete or explicitly de-scoped, automated verification passes for the delivered scope, the user confirms manual verification items, and no critical findings remain.