v0.2.0 — Multi-Agent Orchestration

What's New

selfmodel is a self-evolving multi-AI agent orchestration framework for Claude Code.

Core Features

• 7-Role AI Team — Leader (Opus) + Frontend (Gemini) + Backend (Codex) + Fullstack (Opus Agent) + Researcher + Evaluator + E2E Verifier
• Git Worktree Isolation — Every agent works in a separate worktree. Zero shared state pollution.
• Independent Evaluator — Skeptical prompt, isolated context. Implementer never reviews own code.
• E2E Atomic Verification (v2) — Each acceptance criterion = one atomic verification with evidence.
• AI Slop Auto-Rejection — 10 auto-reject triggers + 8 AI slop pattern penalties.
• Chaos Testing (/rampage) — 7 user personas × 4 surface engines find bugs QA never catches.
• Self-Evolution — Every 10 sprints: measure → diagnose → propose → experiment → evaluate → select.
• Orchestration Loop — Automated plan → dispatch → review → merge → repeat for large projects.

Install

git clone https://github.com/VictorVVedtion/selfmodel.git
cd selfmodel && bash install.sh

Then in Claude Code: /selfmodel:init

Iron Rules

1. Never Fallback — complete solutions, no shortcuts
2. Never Mock — real data only
3. Never Lazy — no TODO, no skipped edge cases
4. Best Taste — naming like prose, architecture worth screenshotting
5. True Artist — every line of code is signed artwork

MIT License