By AskElira
Decompose a software goal into a multi-floor build plan, then dispatch specialized sub-agents (research, builder, architect, validator, fixer) to produce complete, runnable code artifacts in isolated workspaces with cost tracking and automatic retry on failure.
Run the Claude Forge multi-agent build pipeline on a goal. Decomposes into floors, then per floor runs Alba → Vex Gate 1 → David → Vex Gate 2 → Elira → (Steven if rejected) up to 5 iterations, with a Rescue agent as the final fallback.
Manually trigger Steven to fix a specific floor. Reads the workspace state and the floor's failure history (if any), then dispatches Steven to diagnose and patch. Re-validates via Vex Gate 2 and Elira after the fix lands.
Plan a goal into floors without building. Dispatches Elira in PLAN mode and prints the resulting floor decomposition. Useful for sanity-checking a goal before running /claude-forge:build.
Show the status of all Claude Forge workspaces — which goals are in progress, which floors are LIVE, BLOCKED, or pending. Reads ./workspaces/ and any .claude-forge/report.md files.
Research agent. Use this agent at the start of every floor to gather requirements, library choices, code patterns, and architectural context before David builds. Pulls from web search, existing workspace files, and prior knowledge. Produces structured research notes optimized for a builder agent to consume.
Builder agent. Use this agent to write real, working code files into the workspace based on Alba's research. David ships complete files — no stubs, no TODOs, no placeholders. Receives the floor spec, research, and any prior-iteration feedback. Writes files via Write/Edit and reports a build summary.
Architect, planner, and approver. Use this agent at the start of a Claude Forge build to decompose a goal into 3-7 floors, and at the end of each floor to approve or reject David's deliverable. Returns structured JSON in both modes. Call when the user asks to "plan", "design floors", "decompose", "review build", or "approve floor".
Fresh-perspective rescue builder. Use this agent ONLY after 5 failed iterations of David + Steven on the same floor. Receives the full failure history, sees no prior code, and rebuilds from scratch with a deliberately simpler approach. Skips Vex gates and goes straight to Elira for one final review.
Fixer and shipper. Use this agent when a floor has been rejected by Elira or has failed validation, and needs diagnosis + concrete code patches. Also use for runtime errors, missing dependencies, or "it doesn't run" failures. Returns a structured fix plan with patches written directly to the workspace.
Uses power tools
Uses Bash, Write, or Edit tools
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.
Multi-agent autonomous building pipeline for Claude Code. Describe a goal, Claude Forge decomposes it into floors and ships real working code. Reimplemented from the original AskElira/AskElira3 Node prototype as native Claude Code sub-agents — no separate Hermes runtime, no Node infrastructure, no SQLite. The Claude Agent SDK is the runtime.
You say:
/claude-forge:build a REST API for a todo app with SQLite, JWT auth, and tests
Claude Forge:
./workspaces/<slug>/node --check, python3 -m py_compile)All file writes happen through Claude Code's native tools (Write, Edit, Bash). Sub-agents are real Claude Code sub-agents in agents/. Slash commands are real Claude Code slash commands in commands/. No external runtime.
/plugin marketplace add AskElira/claude-forge
/plugin install claude-forge@claude-forge-marketplace
git clone https://github.com/AskElira/claude-forge .claude/plugins/claude-forge
Then reload Claude Code (/reload or restart). The plugin's agents, commands, and skill will appear automatically.
Drop the cloned folder into ~/.claude/plugins/claude-forge/ to use across all your projects.
/claude-forge:build a Python CLI that scrapes Hacker News top stories and emails them daily
/claude-forge:plan a Next.js dashboard with Supabase auth and Stripe billing
/claude-forge:fix workspaces/my-goal-slug 3 "the API returns 500 on /todos POST"
/claude-forge:status
Goal
│
▼
Elira plans floors ──► 3-7 floors with dependsOn
│
▼
Build waves (parallel where deps allow)
│
▼ per floor (max 5 iterations):
│
│ Alba research ─► Vex Gate 1 ─► David build
│ │
│ ▼
│ Syntax + smoke check
│ │
│ ▼
│ Vex Gate 2
│ │
│ ▼
│ Elira approve?
│ ┌──┴──┐
│ YES NO
│ │ │
│ LIVE Steven patches ─► loop
│ │
│ (after 5)
│ ▼
│ Rescue rebuilds
│ ▼
│ Elira final review
│ │
│ LIVE or BLOCKED
claude-forge/
├── .claude-plugin/
│ ├── plugin.json # plugin metadata
│ └── marketplace.json # for git-based marketplace install
├── agents/
│ ├── elira.md # architect, planner, approver
│ ├── steven.md # fixer, debugger
│ ├── alba.md # researcher
│ ├── david.md # builder (writes real files)
│ ├── vex.md # validator (two gates)
│ └── rescue.md # fresh-perspective rebuilder
├── commands/
│ ├── build.md # /claude-forge:build — full pipeline
│ ├── plan.md # /claude-forge:plan — preview floors
│ ├── fix.md # /claude-forge:fix — manual Steven
│ └── status.md # /claude-forge:status — list workspaces
├── skills/
│ └── claude-forge-build/
│ └── SKILL.md # auto-trigger when user asks to "build"
├── examples/
│ └── todo-api.md # sample run-through
├── CLAUDE.md # project guidelines (loaded into context)
├── LICENSE
└── README.md
npx claudepluginhub askelira/claude-forge --plugin claude-forgev9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Unity Development Toolkit - Expert agents for scripting/refactoring/optimization, script templates, and Agent Skills for Unity C# development
Superpowers Plus core skills library for Claude Code: planning, execution routing, TDD, debugging, and collaboration workflows
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review