By mahuebel
Spec-driven design system: visual brainstorming with forge, structured plans with SDD defaults, subagent/agent-team execution with TDD and worktree hygiene
Use when starting a new feature, subsystem, or architecture change — drives visual brainstorming in forge with structured decision records, producing a spec and decisions.json as handoff to forge-sdd:write-plan. Replaces superpowers:brainstorming for projects using forge-sdd.
Execute a forge-sdd plan — subagent dispatch (default) with escalation to agent teams for tagged phases. Integrates native plan mode for the execute gate and per-task stakes-3 approvals. Replaces superpowers:executing-plans for forge-sdd flows.
Author a forge-sdd implementation plan from a spec. Use after brainstorm completes, or invoked with --spec <path> for bring-your-own. Produces plan.md with workstream-first phases, UI skill defaults, and decision cross-references. Replaces superpowers:writing-plans for forge-sdd flows.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A spec-driven design system for Claude Code — visual brainstorming with forge, structured plans with SDD defaults, and subagent/agent-team execution with the TDD, worktree, and commit hygiene of superpowers.
forge-sdd turns the superpowers brainstorm → plan → execute pipeline from prose-only into visual decision-making. Every architecture question becomes a forge topic with three rendered variations. Every accepted variation becomes a structured decision record that survives context compaction. Every plan cross-references the decisions it implements. Every execution phase runs under the right workflow: subagents by default, agent teams where parallel exploration genuinely pays off.
Superpowers gives you a rigorous spec-driven workflow — but every stage is text. For UI and architecture decisions, text is a lossy translation layer. You describe visual intent in prose, Claude interprets, rounds of drift follow:
You: We need to redesign auth. Tokens should rotate, MFA should be optional,
and sessions need to be auditable.
Claude: Here's a plan using JWT cookies with rolling refresh...
You: Actually I meant event-sourced sessions, and MFA might be required
for some tenants, and auditability is a hard compliance requirement
so it needs to be at the storage layer not the app layer...
Claude: [regenerates with partial understanding]
You: Close, but the refresh strategy is wrong, and the topology has
the wrong boundary between the auth service and...
forge solved this for UI (draw variations, click what you like). forge-sdd solves it for everything upstream of UI — architecture, topology, library choices, cross-cutting concerns:
You: /sdd-brainstorm auth redesign
Claude: [creates forge topic "auth-session"]
Topic ready at http://localhost:4546 — 3 approaches:
A: JWT cookies + rolling refresh
B: Server sessions (Redis)
C: Event-sourced sessions (Kafka)
[you open the browser, look at three diagrammed approaches]
[you click "Like" on C]
[you drop a pin on C's MFA box: "must be tenant-configurable"]
[you drop a pin on A's refresh strategy: "use refresh tokens not rolling"]
Claude: ✓ Decision recorded: d-003 — Session storage: event-sourced
Rationale: auditability; tenant-configurable MFA
Addressed: pin-7 (refresh tokens over rolling)
→ Next: Auth — token rotation details
[you drill into the MFA sub-topic, accept a variation]
[continue for data model, deployment topology, API shape…]
Claude: Spec at docs/sdd/specs/2026-04-19-auth-redesign-design.md.
Review and tell me to proceed — I'll advance to plan-writing.
You stay in visual flow through the design phase, and Claude walks out with a durable, structured understanding of what you picked and why.
Three stages, each its own skill, connected by persisted state:
┌──────────────────────┐ ┌──────────────────────┐ ┌──────────────────────┐
│ 1. Brainstorm │ │ 2. Plan │ │ 3. Execute │
│ │ │ │ │ │
│ /sdd-brainstorm │──▶│ /sdd-plan │──▶│ /sdd-execute │
│ │ │ │ │ │
│ forge topics per │ │ fork of superpowers │ │ subagent dispatch │
│ cluster; decision │ │ writing-plans with │ │ (default) + agent │
│ records on accept │ │ UI skill defaults │ │ teams for tagged │
│ │ │ and parallel │ │ phases; native plan │
│ │ │ workstream shape │ │ mode gate at entry │
└──────────┬───────────┘ └──────────┬───────────┘ └──────────┬───────────┘
│ │ │
▼ ▼ ▼
spec.md plan.md worktree branch
decisions.json (commit shas + squash-on-merge
land in-place) + finishing-a-branch
Each stage produces a durable artifact. You can resume from any stage, or start at a later stage with a hand-written input.
curl -fsSL https://bun.sh/install | bash)CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 — only required if any phase in a plan uses execution: team/plugin marketplace add mahuebel/forge-sdd
/plugin install forge-sdd@forge-sdd-marketplace
npx claudepluginhub mahuebel/forge-sdd --plugin forge-sddShared Obsidian knowledge vault with git sync, observation capture, and automatic context injection
Visual workspace for iterating on UI with Claude Code — point, comment, react, and refine in real time instead of describing changes in prose
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Memory compression system for Claude Code - persist context across sessions
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Write feature specs, plan roadmaps, and synthesize user research faster. Keep stakeholders updated and stay ahead of the competitive landscape.