Getting Started

Install & configure effectum

effectum takes about 2 minutes to install. Once set up, write a spec and Claude builds it — with tests, quality gates, and guardrails.

Prerequisites

  • Claude Code installed and configured
  • Node.js 18+ and npx available
  • A project directory (new or existing)

Installation

The recommended way to install effectum is the interactive installer. It walks you through selecting a stack preset and scope.

terminal
npx @aslomon/effectum

Installation options

GlobalRecommended for individuals

Installs to ~/.claude/ — available in all projects. Ideal if you use Claude Code across many repositories.

terminal
npx @aslomon/effectum --global
LocalRecommended for teams

Installs to ./.claude/ — committed with your repo so everyone shares the same workflow and guardrails.

terminal
npx @aslomon/effectum --local
Non-interactiveFor CI / automation

Skips all prompts. Uses Claude Code runtime defaults. Good for scripted setups and CI pipelines.

terminal
npx @aslomon/effectum --global --claude
Classic git cloneFull control

Clone the repo and use /setup to configure manually. Best when you want to inspect every file before applying.

terminal
git clone https://github.com/aslomon/effectum.git
cd effectum && claude
/setup ~/my-project

What gets installed

10 workflow commands

Slash commands for the entire lifecycle — planning, TDD, verification, code review, autonomous building

PRD Workshop

8 knowledge files + 12 commands for guided specification writing with adaptive questioning

Quality gates

8 automated checks — build, types, lint, tests, security, debug logs, type safety, file size

Stack preset

Pre-configured CLAUDE.md, settings.json, and guardrails tuned for your technology stack

Guardrails

Safety rules that prevent common mistakes — protected files, blocked destructive commands, architecture rules

4 MCP servers

Context7 (docs), Playwright (E2E), Sequential Thinking, Filesystem — pre-configured and ready

Your first feature

Once installed, the typical workflow is four steps. You write a spec, Claude builds it.

terminal
# Open your project in Claude Code
cd ~/my-project && claude

# Write a specification (guided mode)
/prd:new

# Review the plan before building
/plan docs/prds/001-my-feature.md
# → Claude creates implementation plan
# → You approve or adjust

# Build with test-driven development
/tdd

# Verify all quality gates pass
/verify

For well-defined features, use /prd:express for a one-shot spec. Use /prd:new for guided discovery when the idea is still vague.

Autonomous overnight builds

For complex features with a thorough spec, use the Ralph Loop. Claude iterates autonomously — writing code, running tests, fixing errors — until every quality gate passes.

terminal
/ralph-loop "Build the auth system per PRD"
  --max-iterations 30
  --completion-promise "All tests pass, build succeeds, 0 lint errors"

The completion promise is only output when it is 100% true. Claude cannot lie to exit the loop. At 80% of max iterations, it writes a status report of what is done and what remains.

Choosing your autonomy level

Configure how much Claude decides independently during /setup. Change anytime in .claude/settings.json.

LevelAsks beforeBest for
ConservativeMost changesTeams, learning
StandardAmbiguous specsDaily development
Full AutonomyAlmost nothingOvernight builds