Usage Guide

Complete guide to all 6 Forge DevKit modules — what they do, how to use them, and why they matter.

All commands run inside your AI session (Claude Code, Cursor, or Claude Desktop). Forge DevKit is a plugin system — commands become available after installation.

forge-core

Core+

The foundation. One command makes your AI understand your entire project.

forge-core runs a 7-gate architecture audit that auto-detects your stack, layers, conventions, and patterns. It generates project-specific dev-skills and 50+ quality guardrails that your AI follows in every session. No manual prompt engineering. No re-explaining your architecture.

What you get

  • Auto-detects stack, architecture patterns, and conventions
  • Generates 50+ quality patterns tailored to your project
  • Creates project-specific dev-skills your AI loads automatically
  • Scoped pipeline phases prevent AI from going off-track
  • 15 LLM rationalization detectors catch AI shortcuts
/forge:setup Setup wizard, ecosystem dashboard, upgrade, and reconfigure.
/forge:setup Run the 7-gate wizard (first time) or show the ecosystem dashboard (after setup)
/forge:setup upgrade --check Check for plugin updates
/forge:setup upgrade Apply available updates
/forge:setup reconfigure Re-run specific gates without full setup
/forge:patch Quick config patches without running the full wizard.
/forge:patch Interactive patch mode — select what to adjust
/dev Start a development task using the generated pipeline.
/dev Begin a feature/fix using your project's dev-skills and quality patterns
/forge:sync Sync generated artifacts after architecture changes
/forge:close Finalize and close the current development task

Typical workflow: Install → /forge:setup (7-gate audit) → /dev to start building. Run /forge:setup anytime for dashboard.

forge-product

Pro+

Describe features in business language. AI builds from spec, not guesswork.

forge-product bridges the gap between what you want and what AI builds. Describe a feature in plain language — it generates user flows, user stories with acceptance criteria, and UX specifications. These artifacts feed directly into the development pipeline, so your AI codes from verified requirements instead of hallucinated assumptions.

What you get

  • 9 product archetypes (SaaS, marketplace, API platform, etc.)
  • Generates user flows, stories, and acceptance criteria automatically
  • 50+ UX criteria library for consistent quality
  • Stories feed directly into /dev pipeline — zero copy-paste
  • Supports multiple methodologies (Lean, Jobs-to-be-Done, etc.)
/forge:product-setup Configure product design module — select archetype, methodology, integrations.
/forge:product-setup Interactive setup wizard for product module
/forge:product Product design operations — features, epics, roadmap, strategy.
/forge:product design Design a feature with user flows, stories, and acceptance criteria
/forge:product epic Create an epic with breakdown into stories
/forge:product roadmap Generate or update a product roadmap
/forge:product strategy Define or refine product strategy (positioning, ICP, GTM)

Typical workflow: /forge:product-setup → /forge:product design for each feature → stories feed into /dev pipeline.

forge-qa

Pro+

Tests that trace back to requirements. Not AI guesswork.

forge-qa generates tests from acceptance criteria, not from AI assumptions. It supports 10 testing modes and maintains a 4-level traceability matrix: acceptance criteria → unit tests, use cases → E2E tests, UX criteria → visual tests, and LLM-as-Judge for AI output quality. Every test proves a requirement — no orphan tests, no missing coverage.

What you get

  • Generates tests from acceptance criteria — not guesswork
  • 4-level traceability: AC → unit, UC → E2E, UX → visual, Judge
  • 10 testing modes for different scenarios
  • LLM-as-Judge evaluates AI-generated output quality
  • Quality gates block merges until requirements are covered
/forge:qa-setup Configure QA module — select test frameworks, coverage thresholds, traceability level.
/forge:qa-setup Interactive QA configuration wizard
/forge:qa Test operations — generation, coverage analysis, quality gates.
/forge:qa generate Generate tests from acceptance criteria or user stories
/forge:qa coverage Analyze test coverage and suggest missing tests
/forge:qa trace Show traceability matrix (AC → test → code)
/forge:qa judge Run LLM-as-Judge to evaluate test quality
/forge:qa gate Run quality gate checks before merge

Typical workflow: /forge:qa-setup → /forge:product design (generates AC) → /forge:qa generate → /forge:qa gate before merge.

forge-tracker

Pro+

Your backlog lives in your tracker. Now your AI reads it too.

forge-tracker connects your existing task management tool to the AI development pipeline. Pick a task from GitHub Issues, Linear, ClickUp, Notion, or Jira — your AI gets the full context (description, acceptance criteria, labels, priority) and starts working. When done, it syncs status and links the PR back to the task. No more copy-pasting issue descriptions into prompts.

What you get

  • 5 integrations: GitHub Issues, Linear, ClickUp, Notion, Jira
  • Pick tasks directly from your tracker — AI gets full context
  • Auto-syncs task status and progress back to tracker
  • Links PRs and commits to tasks automatically
  • AC validation ensures task requirements are met before closing
/forge:tracker-setup Configure task tracker integration — select provider and connect.
/forge:tracker-setup Interactive tracker setup (auto-detects available integrations)
/forge:tracker Task management operations — pick, sync, close tasks.
/forge:tracker pick Pick a task from your tracker to work on
/forge:tracker sync Sync task status and progress back to tracker
/forge:tracker close Close task with summary and link to PR/commit

Typical workflow: /forge:tracker-setup → /forge:tracker pick → /dev → /forge:tracker sync → /forge:tracker close.

forge-autopilot

Bundle

Run 3 agents in parallel. Zero merge conflicts.

forge-autopilot orchestrates autonomous AI teammates like Kubernetes orchestrates containers. Give it a list of features — it triages them, detects potential conflicts (shared models, overlapping files), groups non-conflicting features into parallel waves, and deploys agents to isolated worktrees. Each agent follows your full pipeline (guardrails, tests, quality gates). When a wave completes, it auto-merges and starts the next wave.

What you get

  • Backlog triage classifies features by complexity and risk
  • Conflict detection prevents agents from editing the same files
  • Wave execution runs non-conflicting features in parallel
  • 3 autonomy modes: supervised, semi-auto, full-auto
  • Auto-merge with quality gate verification before merge
/forge:autopilot Launch batch autonomous development pipeline.
/forge:autopilot --setup Configure autonomy mode and feature sources
/forge:autopilot Start batch with configured sources (triage → waves → execute)
/forge:autopilot "auth, payments" Start batch with specific features
/forge:autopilot --resume Resume interrupted batch from last checkpoint
/forge:autopilot --abort Abort current batch and clean up worktrees

Typical workflow: /forge:autopilot --setup → /forge:autopilot → agents triage, detect conflicts, execute in waves → review results.

forge-worktree

Bundle

Isolated branches with zero port conflicts. Automatic cleanup.

forge-worktree manages the git worktree lifecycle for parallel development. Each worktree gets its own directory, Docker port offsets, and environment configuration — no more port conflicts between branches. Three templates cover common setups: Docker backend (with port isolation), Frontend (with dev server offsets), and Generic. Worktrees are auto-provisioned by forge-autopilot or created manually.

What you get

  • 3 templates: Docker backend, Frontend, Generic
  • Automatic Docker port isolation per worktree
  • Auto-cleanup when branches are merged
  • Integrates with forge-autopilot for autonomous provisioning
  • Zero manual configuration for environment isolation
/forge:worktree-setup Configure worktree templates and isolation settings.
/forge:worktree-setup Interactive worktree configuration — select template (Docker backend, Frontend, Generic)

Typical workflow: /forge:worktree-setup → worktrees are auto-provisioned by forge-autopilot or created manually via git worktree.

Feedback & Bug Reports

Found a bug? Have a suggestion? Use the built-in feedback command — it sends structured reports directly to the Forge team.

/forge:feedback Send feedback, bug reports, or feature requests.
/forge:feedback Interactive feedback wizard — select type (bug, friction, suggestion), describe the issue, auto-attaches session context

The feedback command collects:

  • Your description — what happened, what you expected
  • Session context — plugin versions, detected stack, execution mode (auto-attached, no manual work)
  • Feedback type — friction (something was hard), bug (something broke), suggestion (feature request)

Reports are sent securely using your license key. No source code or file contents are included — only metadata and your description.

Alternatively, email support@reumbra.com for direct support.

Ready to get started?

One-time purchase. 14-day money-back guarantee.

Get Forge →