自动代理工厂 是 AI Skill Hub 本期精选n8n工作流之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
自动代理工厂 是一款基于 JavaScript 开发的开源工具,专注于 ai-agents、human-in-the-loop、llmops 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
自动代理工厂 是一款基于 JavaScript 开发的开源工具,专注于 ai-agents、human-in-the-loop、llmops 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
# 方式一:npm 全局安装 npm install -g auto-agent-factory # 方式二:npx 直接运行(无需安装) npx auto-agent-factory --help # 方式三:项目依赖安装 npm install auto-agent-factory # 方式四:从源码运行 git clone https://github.com/fangwendongcs/Auto-agent-factory cd Auto-agent-factory npm install npm start
# 命令行使用
auto-agent-factory --help
# 基本用法
auto-agent-factory [options] <input>
# Node.js 代码中使用
const auto_agent_factory = require('auto-agent-factory');
const result = await auto_agent_factory.run(options);
console.log(result);
// n8n 工作流配置步骤 // 1. 在 n8n 中点击 "Import Workflow" // 2. 粘贴 JSON 文件内容或上传文件 // 3. 配置必要的 Credentials: // - Settings → Credentials → New // - 选择对应服务类型填写 API Key // 4. 激活工作流 (Toggle ON) // 5. 通过 Webhook 或定时触发器运行
A mock-first, goal-driven n8n workflow skeleton for bounded, testable, and human-reviewable AI Agent automation.
Technical positioning: Goal-Driven Agent Workflow with n8n.
Language: English | 简体中文
Auto Agent Factory is a mock-first AI Agent workflow skeleton built with n8n. It turns an agent request into a bounded workflow contract: define a goal, define success criteria, run a controlled executor step, check the result, and decide whether to finish, revise, stop, or require human review.
Current stage: v0.7 / Human-in-the-loop Controlled Execution Boundary Verified. The project has validated mock, dry-run, real-readonly stub routing, one OpenAI-compatible read-only provider sandbox call, criterion-indexed evidence alignment, and approval-boundary decisions for read-only, high-risk write-like, and forbidden action classes.
This is not a production autonomous agent. The verified real provider path is read-only and returns needs_review; it does not execute real Codex/coding-agent tasks, shell commands, file writes, Git modifications, external write actions, or live SaaS user workflows.
Current design target: V0.9b local audit report artifact option, still no production autonomous execution.
The V0.4 preparation checklist is tracked in docs/V0_4_PROVIDER_INTEGRATION_PREP.md. The first provider interface decision is recorded in docs/ADR_0001_REAL_READONLY_PROVIDER_SELECTION.md. The minimal implementation plan is tracked in docs/V0.4C_REAL_READONLY_IMPLEMENTATION_PLAN.md. The evaluator-quality phase is documented in docs/V0.6_EVALUATOR_QUALITY_PLAN.md. The controlled-execution boundary design is tracked in docs/V0.7_CONTROLLED_EXECUTION_BOUNDARIES.md. The staging pilot, audit, and rollback design is tracked in docs/V0.8_STAGING_PILOT_AUDIT_ROLLBACK_DESIGN.md. The sanitized audit log prototype is tracked in docs/V0.8B_SANITIZED_AUDIT_LOG_PROTOTYPE.md. The dev-only audit storage plan is tracked in docs/V0.8C_DEV_ONLY_AUDIT_STORAGE_PLAN.md. The dev-only JSONL audit storage prototype is tracked in docs/V0.8D_DEV_ONLY_JSONL_AUDIT_STORAGE_PROTOTYPE.md. The dev-only audit CLI is tracked in docs/V0.8E_DEV_ONLY_AUDIT_CLI.md. The staging replay closeout is tracked in docs/V0.8F_STAGING_REPLAY_CLOSEOUT.md. The audit review report generator is tracked in docs/V0.9_AUDIT_REVIEW_REPORT_GENERATION.md. The dev-only local report artifact option is tracked in docs/V0.9B_LOCAL_AUDIT_REPORT_ARTIFACT.md.
Install dependencies:
npm install
Run tests:
npm test
Validate workflow JSON:
npm run workflow:validate:all
Run dry-run deployment check:
npm run workflow:dry-run
Check n8n import readiness:
npm run import:check
Optional smoke-test payload generation:
npm run smoke:goal-driven
| Module | File | Responsibility | Current status |
|---|---|---|---|
| Goal-Driven Master Workflow | workflows/goal_driven_master.workflow.json | Receives goal payloads, validates input, initializes run/task IDs, dispatches executor/checker, routes final response. | Implemented |
| Agent Task Executor Workflow | workflows/agent_task_executor.workflow.json | Executes one bounded task iteration and returns a normalized agent_result. | Implemented with mock, dry-run, and real-readonly stub modes |
| Criteria Checker Workflow | workflows/criteria_checker.workflow.json | Evaluates executor evidence against criteria and returns pass/fail/unknown checks. | Implemented |
| Goal-Driven Error Handler Workflow | workflows/error_handler.workflow.json | Handles failed workflow executions and produces recovery context. | Implemented |
高效的自动化工作流工具包
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
经综合评估,自动代理工厂 在n8n工作流赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | Auto-agent-factory |
| 原始描述 | 开源n8n工作流:A production-ready toolkit to accelerate and automate the end-to-end lifecycle o。⭐10 · JavaScript |
| Topics | ai-agentshuman-in-the-loopllmopsn8nworkflow-automationjavascript |
| GitHub | https://github.com/fangwendongcs/Auto-agent-factory |
| License | MIT |
| 语言 | JavaScript |
收录时间:2026-05-29 · 更新时间:2026-05-30 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端