AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻譯準備中
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證⚡ The Universal AI Coding Agent Toolkit & OS. Sync rules across Cursor, Claude Code, Windsurf, Antigravity, Copilot, Roo & Aider + AST context compressor + rule linter + 7 built-in agent skills.
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
⚡ AgentForge
The Universal AI Coding Agent Toolkit & Operating System
Single Source of Truth for all your AI Coding Agents + 85% Token AST Compressor + Rule Linter + 7 Built-in Agent Skills.檢視完整 README ↗ █████╗ ██████╗ ███████╗███╗ ██╗████████╗███████╗ ██████╗ ██████╗ ██████╗ ███████╗
██╔══██╗██╔════╝ ██╔════╝████╗ ██║╚══██╔══╝██╔════╝██╔═══██╗██╔══██╗██╔════╝ ██╔════╝
███████║██║ ███╗█████╗ ██╔██╗ ██║ ██║ █████╗ ██║ ██║██████╔╝██║ ███╗█████╗
██╔══██║██║ ██║██╔══╝ ██║╚██╗██║ ██║ ██╔══╝ ██║ ██║██╔══██╗██║ ██║██╔══╝
██║ ██║╚██████╔╝███████╗██║ ╚████║ ██║ ██║ ╚██████╔╝██║ ██║╚██████╔╝███████╗
╚═╝ ╚═╝ ╚═════╝ ╚══════╝╚═╝ ╚═══╝ ╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚═╝ ╚═════╝ ╚══════╝
Single Source of Truth for all your AI Coding Agents + 85% Token AST Compressor + Rule Linter + 7 Built-in Agent Skills.
Quickstart • Features • Comparison • Web Studio • Skills Library
Developers using AI coding assistants are trapped in fragmented configurations:
.cursorrules, .cursor/rules/*.mdc, CLAUDE.md, .windsurfrules, .agents/AGENTS.md, .github/copilot-instructions.md, and .aider.conf.yml.AgentForge acts as the Universal Operating System for your AI workflow:
.agentforge.yaml $\to$ automatically compiles to all 7+ agent formats.tdd-architect, security-auditor, minimalist-architect, fastapi-master, etc.).Note: AgentForge is not published on npm yet, so
npx agentforge/npm install -g agentforgewill not work. Run it from GitHub instead (Node.js 18+, npm 10+ recommended).
Run directly from GitHub with npx (it builds on install):
# 1. Initialize AgentForge in any project (auto-detects language & framework)
npx github:mrzroot/agentforge init
# 2. Compile & Synchronize rules across all AI agents
npx github:mrzroot/agentforge sync
# 3. Compress codebase into a token-optimized context bundle
npx github:mrzroot/agentforge pack --out CONTEXT.md
# 4. Audit rules & prompt health
npx github:mrzroot/agentforge lint
# 5. Launch interactive Web Dashboard
npx github:mrzroot/agentforge ui
Or install from source and link the agentforge command globally:
git clone https://github.com/mrzroot/agentforge.git
cd agentforge
npm install # also builds via the prepare script
npm link # `agentforge` (and `af`) are now on your PATH
| Feature | Manual Rules | Cursor Only | Claude Only | AgentForge ⚡ |
|---|---|---|---|---|
| Multi-Agent Sync (Cursor, Claude, Windsurf, Antigravity, Copilot, Aider, Cline) | ❌ | ❌ | ❌ | ✅ Full Support |
| AST Token Compression (80-90% token reduction) | ❌ | ❌ | ❌ | ✅ Built-in |
| Secret & API Key Scanner in Rules | ❌ | ❌ | ❌ | ✅ Built-in |
| Prompt Conflict & Bloat Linter | ❌ | ❌ | ❌ | ✅ Built-in |
| Curated Skills Library | ❌ | ❌ | ❌ | ✅ 7 Built-in Skills |
| Local Visual Web Studio UI | ❌ | ❌ | ❌ | ✅ Zero-Config |
agentforge syncCompiles your single source of truth (.agentforge.yaml) into targeted formats:
.cursorrules & .cursor/rules/project-rules.mdcCLAUDE.md.windsurfrules.agents/AGENTS.md.github/copilot-instructions.md.aider.conf.yml & .aider.prompt.md.clinerulesagentforge sync --targets cursor claude windsurf
agentforge pack (Token-Killer Engine)Extracts structural interfaces, function signatures, types, and docstrings while stripping bulky routine bodies.
agentforge pack --out PROJECT_CONTEXT.md
Token Benchmark:
48,200 tokens (~$0.24 per prompt)6,150 tokens (~$0.03 per prompt)agentforge lintCalculates an automated Agent Health Score (0 - 100%):
rm -rf /).agentforge skills add <name>Installs battle-tested personas and skills directly into your agent instructions:
agentforge skills list
agentforge skills add tdd-architect
agentforge skills add security-auditor
agentforge skills add minimalist-architect
agentforge uiLaunches the interactive Web Studio on http://localhost:3737:
minimalist-architect: Enforces minimalist senior dev philosophy (YAGNI, standard library first, zero bloat).tdd-architect: Strict Test-Driven Development (Red-Green-Refactor) with boundary isolation.security-auditor: OWASP Top 10 hardening, input sanitization, and SQL injection prevention.token-optimizer: Context window management and prompt payload compression.fastapi-master: Pydantic v2 schemas, async route handlers, and dependency injection.react-nextjs-pro: React Server Components (RSC), App Router, Server Actions, and performance.docker-devops: Multi-stage, distroless, non-root hardened Dockerfiles.Contributions are warmly welcomed! Feel free to submit an issue, open a pull request, or add new skills to the registry.
git clone https://github.com/mrzroot/agentforge.git
cd agentforge
npm install
npm run build
npm test
MIT © MRZ - Built with ❤️ for AI Agent Developers worldwide.