rpa

RPA - Research, Plan, Act

A collection of Claude Code slash commands for structured software development workflows. These commands help you systematically research codebases, create implementation plans, and execute them with proper verification.

Philosophy

The RPA workflow encourages a methodical approach to software development:

1. Research - Understand the codebase before making changes
2. Plan - Create detailed, phased implementation plans with clear success criteria
3. Act - Implement the plan with continuous verification

This approach reduces errors, improves code quality, and creates documentation that helps with future maintenance.

Commands Overview

Command	Description
/research_codebase	Document and explain the codebase as it exists
/create_plan	Create detailed implementation plans through interactive iteration
/implement_plan	Execute approved plans with verification at each phase
/iterate_plan	Update existing plans based on feedback
/enhance_plan	Synthesize multiple opinions into plan improvements
/enhance_research	Improve research documents with additional findings
/validate_plan	Verify that a plan was correctly implemented
/create_handoff	Create handoff documents to transfer work between sessions
/resume_handoff	Resume work from a handoff document
/commit	Create well-structured git commits for session changes
/debug	Investigate problems via logs, DB state, and git history without editing files
/create_test_plan	Create comprehensive test plans and TDD strategies
/tdd	Execute full TDD cycle with Red/Green/Refactor verification
/test_suite	Create, update, and maintain test suites (audit, init, update, gaps, run, ci, standardize)
/refactor_candidates	Discover and index God-like modules as refactoring candidates
/refactor	Refactor monolithic modules into focused, testable modules
/tech_debt_sweep	Scan codebase for technical debt and generate paydown plan
/tech_debt_trends	Analyze technical debt trends over time
/aidlc_init	Bootstrap the current repository for the AI-DLC compatibility workflow
/aidlc_start	Initialize or resume the AI-DLC compatibility workflow
/aidlc_inception	Execute approved inception stages and generate canonical artifacts
/aidlc_bolt	Execute one construction unit in the AI-DLC workflow
/aidlc_build_test	Run the global Build and Test stage
/aidlc_operations	Generate experimental operations artifacts
/aidlc_feedback	Capture experimental retrospective feedback

When to Use Plugin vs Native Claude Code

Claude Code ships native skills and agents that partially overlap with this plugin. Use the right tool for the altitude of the task:

Task	Use native	Use this plugin
Clean up a diff (naming, dead code, small simplifications)	/simplify	—
Find bugs in a branch/PR diff	/code-review	—
Check an implementation against its plan and success criteria	—	/validate_plan (then follow up with /code-review)
Decompose a God module into focused modules (multi-phase, with API snapshots and rollback)	—	/refactor, /refactor_candidates
Ad-hoc "where is X / how does Y work" exploration	Explore agent	—
Durable research docs in thoughts/shared/research/	—	/research_codebase
Security audit of pending changes	/security-review	—
Find hardcoded config that should be externalized	—	config-auditor (via /tech_debt_sweep)
Deep multi-source web research report	deep-research	—
Targeted web lookups inside plan/research flows	—	web-search-researcher (spawned by /create_plan, /research_codebase)

Rule of thumb: native skills work at the diff level on the current branch; this plugin works at the workflow level — persistent artifacts in thoughts/, multi-phase orchestration, and cross-session state.

Installation

Quick Install (Recommended)

Copy commands, agents, and scripts to your global Claude configuration:

# Create directories if they don't exist
mkdir -p ~/.claude/commands
mkdir -p ~/.claude/agents
mkdir -p ~/.claude/scripts
mkdir -p ~/.claude/hooks

# Copy commands
cp commands/*.md ~/.claude/commands/

# Copy agents (enables parallel sub-agents)
cp agents/*.md ~/.claude/agents/

# Optional: copy the compatibility bootstrap if using AI-DLC workflows
cp CLAUDE.md ~/.claude/CLAUDE-rpa-aidlc.md

# Copy and make scripts executable
cp scripts/*.sh ~/.claude/scripts/
chmod +x ~/.claude/scripts/*.sh

# Optional: hooks for deterministic quality gates — merge the "hooks" object
# from hooks/hooks.json into ~/.claude/settings.json (hooks are read from
# settings, not from a standalone file; see "How Hooks Work" below)

Plugin Install (Alternative)