Agentic AI Atlas

Wiki article

library/rpikit

Reading · 3 min

RPIKit Methodology (Library) reference

RPIKit implements the Research-Plan-Implement (RPI) framework for disciplined AI-assisted software development. It enforces structured progression through distinct phases with human approval checkpoints, stakes-based rigor scaling, and mandatory verification at every step.

Page nodewiki/library/rpikit.mdNearby pages · 111Documents · 1

Continue reading

Nearby pages in the same section.

Documented graph nodes

Records linked directly from this page’s Page node.

RPiKit · Specialization

RPIKit Methodology

**Source**: bostonaholic/rpikit by Matthew Boston **Category**: Structured Software Development / Research-Plan-Implement Framework **License**: See upstream repository

Overview

Core Principles

**Iron Law**: "Do NOT explore the codebase until the problem is understood"
**Stakes-Based Rigor**: Planning intensity scales with change risk (low/medium/high)
**Test-First**: Every code-changing task specifies tests before implementation
**Verification-Before-Completion**: Never declare done without running verification
**Human Approval Gates**: Explicit human approval between major phases
**Soft-Gating Reviews**: Code and security reviews preserve user autonomy

Process Files

Process	File	Description	Task Count
Brainstorm	`rpikit-brainstorm.js`	Clarify vague requirements through exploration	3
Research	`rpikit-research.js`	Systematic codebase exploration (Iron Law)	5
Plan	`rpikit-plan.js`	Stakes-based implementation planning with test-first	7
Implement	`rpikit-implement.js`	Disciplined plan execution with verification	7
Review	`rpikit-review.js`	Combined code quality and security review	4
Decision	`rpikit-decision.js`	Architecture Decision Records (ADRs)	3

Skills Catalog

Skill	Directory	Description
codebase-research	`skills/codebase-research/`	Systematic codebase exploration (Iron Law)
plan-writing	`skills/plan-writing/`	Stakes-based implementation planning
plan-implementation	`skills/plan-implementation/`	Disciplined plan execution with verification
code-review	`skills/code-review/`	Structured code quality assessment
security-review	`skills/security-review/`	Security vulnerability assessment
brainstorming	`skills/brainstorming/`	Requirement clarification for vague goals
decision-documentation	`skills/decision-documentation/`	Architecture Decision Records
test-driven-development	`skills/test-driven-development/`	Test-first development practice
systematic-debugging	`skills/systematic-debugging/`	Hypothesis-driven defect investigation
verification	`skills/verification/`	Verification-before-completion discipline
finishing-work	`skills/finishing-work/`	Final completion and summary protocol

Agents Catalog

Agent	Directory	Role
file-finder	`agents/file-finder/`	File Discovery Specialist
web-researcher	`agents/web-researcher/`	External Context Gatherer
code-reviewer	`agents/code-reviewer/`	Code Quality Assessor
security-reviewer	`agents/security-reviewer/`	Security Vulnerability Assessor
test-runner	`agents/test-runner/`	Test Execution Specialist
verifier	`agents/verifier/`	Implementation Verifier
debugger	`agents/debugger/`	Defect Investigator

Workflow Lifecycle

Code

[Optional] Brainstorm ("what to build") -> Research (Iron Law) -> Plan (stakes-based) -> Implement (verify each step) -> Review (code + security) -> Done
                                                                                                     ^                              |
                                                                                                     +--- Decision (ADR at any point) ---+

Cross-cutting concerns applied throughout:

verification - Every step verified before proceeding
test-driven-development - Tests specified before implementation
systematic-debugging - Failures investigated with hypothesis-driven methodology
finishing-work - Completion confirmed against success criteria

Stakes Classification

Level	Profile	Planning Approach
Low	Isolated, reversible, minimal impact	Brief plan, inline if needed
Medium	Multiple files, moderate scope, testable	Standard RPI cycle
High	Architectural, difficult rollback, security-sensitive	Comprehensive research and detailed planning required

Output Artifacts

docs/plans/YYYY-MM-DD-<topic>-research.md - Research findings
docs/plans/YYYY-MM-DD-<topic>-plan.md - Implementation plans
docs/decisions/NNNN-<title>.md - Architecture Decision Records

Philosophy

**Understand before acting** - Research before planning
**Plan before coding** - Plans before implementation
**Implement with discipline** - Step-by-step verification
**Scale rigor to risk** - Stakes-based planning depth
**Preserve autonomy** - Reviews are soft gates, humans decide