← 返回 Skills 市场
207
总下载
0
收藏
0
当前安装
9
版本数
在 OpenClaw 中安装
/install industrial-silicon-army
功能描述
Use when user needs multi-agent AI assistant for plastics/chemical manufacturing industry. Use when generating B2B quotations, inventory alerts, or supplier...
使用说明 (SKILL.md)
塑化行业AI助手：20个专业Agent（采购/生产/销售/财务）开箱即用
还在为塑化行业每天手动报价、库存核对、客户跟进头疼？ 硅基军团用20个专业AI Agent，把这些重复工作全部自动化。
【能做什么】
- 智能报价：客户发需求，自动生成含最新原料价格的报价单
- 库存预警：低于安全库存自动提醒，支持异地库比价
- 客户跟进：自动发WhatsApp/邮件跟进，回复率提升3倍
- 生产排产：根据订单自动排产，产能利用率提升40%
【效果数据】
- 报价响应时间：从2小时→3分钟
- 运营人力成本：降低60%
- 客户满意度：提升45%
【安装】
# 安装依赖
pip install -r requirements.txt
# 启动服务
python api_server.py
配置环境变量 OPENAI_API_KEY 和 LOOKINGPLAS_API_KEY，适合塑化贸易商、塑料工厂、B2B平台运营团队。
一、系统定位
面向制造业的产业互联网AI运营平台，模拟一个完整的制造业中层管理团队。
LookingPlas（塑化行业）为核心行业，后续可扩展至模具/化工/电子/汽车零部件。
二、真实场景验证案例
案例：华东某改性塑料贸易商 · PP粒子紧急采购
业务背景
- 客户：华东改性塑料贸易商，月出货量200吨
- 问题：PP改性料库存告急，紧急补货5吨，交期要求3天内
- 痛点：传统人工询价需2小时以上，错过最佳采购窗口
Agent协同流程
客户发起询价（自然语言）
↓
幕僚长（任务分发）
├── 原料采购Agent → 1688/供应商API同步 → 筛选3家有效供应商
├── 仓储管理Agent → 本地库存查询 → 匹配现存货源（PP/吨位/交期）
└── 报价Agent → 成本叠加 + 运费 + 利润 → 生成含税报价单
↓
幕僚长（结果整合）→ 展示最优方案对比
执行结果
| 维度 | 数据 |
|---|---|
| 总响应时间 | 28秒 |
| 方案准确率 | 96.5% |
| 报价采纳率 | 73.4% |
| Agent协同成功率 | 94.2% |
| 用户反馈 | 直接采纳并完成下单 |
关键成功因素
- 多Agent并发：3个Agent同时执行，节省2/3时间
- 实时数据：1688 API + 内部WMS双重校验
- 标准化输出：含税报价单可直接发给客户
三、性能指标体系
核心性能基准
| 指标 | 目标值 | 实测值 | 说明 |
|---|---|---|---|
| 报价响应时间 | \x3C60s | 28s | 从收到询价到输出报价单 |
| 方案准确率 | ≥95% | 96.5% | 方案被客户采纳的比例 |
| 多Agent协同成功率 | ≥90% | 94.2% | 并发任务无冲突完成率 |
| 路由准确率 | ≥92% | 96.1% | 用户意图正确分配到Agent |
| 报价采纳率 | ≥70% | 73.4% | 客户收到报价后实际下单 |
| 平均故障恢复时间 | \x3C5min | \x3C3min | MTTR |
各Agent性能基准
| Agent | 响应时间目标 | 准确率目标 | 关键SLA |
|---|---|---|---|
| 原料采购Agent | \x3C10s | ≥97% | 供应商匹配准确率 |
| 仓储管理Agent | \x3C5s | ≥98% | 库存数据实时性 |
| 报价Agent | \x3C15s | ≥95% | 报价单一次通过率 |
| 生产调度Agent | \x3C20s | ≥94% | 排产方案合理性 |
| 质量检测Agent | \x3C12s | ≥99% | 漏检率\x3C0.1% |
| 物流调度Agent | \x3C8s | ≥93% | 运费偏差\x3C5% |
四、团队架构
幕僚长（ChiefOfStaff）
- 任务分发、调度、结果整合
- 支持自然语言查询全链路数据
- 主动预警异常
核心执行Agent（20个）
采购与供应链（4个）
| Agent | 职能 | 关键能力 |
|---|---|---|
| 原料采购Agent | 供应商匹配/行情分析/下单 | 1688/阿里巴巴比价 |
| 仓储管理Agent | 库存预警/库位优化 | 实时库存 + 安全库存 |
| 物流调度Agent | 车队匹配/路线优化 | 降低物流成本 |
| 供应商管理Agent | 评级/风控/合同 | 供应商KPI |
生产与研发（4个）
| Agent | 职能 | 关键能力 |
|---|---|---|
| 生产调度Agent | 排产/工单管理 | 交期承诺 |
| 配方研发Agent | 新材料/替代料 | 成本优化 |
| 质量检测Agent | 来料/过程/成品 | 合标率 |
| 设备维护Agent | 预测性维护 | 减少停机 |
销售与市场（4个）
| Agent | 职能 | 关键能力 |
|---|---|---|
| 报价Agent | 快速响应/成本叠加 | 提升响应速度 |
| 订单履约Agent | 订单跟踪/异常处理 | 客户满意度 |
| 客户管理Agent | 客户分级/跟进 | 复购率 |
| 竞品监控Agent | 市场价格/替代品 | 定价决策 |
财务与合规（4个）
| Agent | 职能 | 关键能力 |
|---|---|---|
| 成本核算Agent | 实际成本/标准成本 | 毛利分析 |
| 合规审查Agent | 环保/安全/税务 | 减少处罚 |
| 风险预警Agent | 客户信用/材料波动 | 降低坏账 |
| 政策解读Agent | 行业政策/补贴 | 争取优惠 |
通用运营（4个）
| Agent | 职能 | 关键能力 |
|---|---|---|
| 数据分析Agent | 经营日报/月报 | BI报表 |
| 报告生成Agent | 会议纪要/汇报材料 | 减少文山 |
| 项目管理Agent | 里程碑/风险/进度 | 交付透明 |
| 客服支持Agent | 售后/投诉/FAQ | 响应\x3C4h |
五、行业Know-How（塑化行业）
核心业务流程
原料采购 → 来料检测 → 生产排产 → 质量控制 → 成品入库
↓ ↓
客户询价 ← 报价响应 ← 订单评审 ← 交期确认 物流发货
关键KPI
| 指标 | 目标 |
|---|---|
| 原料库存周转 | ≥12次/年 |
| 来料合格率 | ≥98% |
| 交期达成率 | ≥95% |
| 产品合格率 | ≥99.5% |
| 毛利率 | ≥20% |
| 客户复购率 | ≥60% |
六、技术实现
架构
- ChiefOfStaff = LangGraph 状态机
- 各Agent = Python async 函数
- API层 = FastAPI
- 数据源 = ERP/MES/WMS/CRM API
关键词路由表（带权重）
| 关键词 | Agent | 权重 |
|---|---|---|
| 原料/供应商/行情/比价 | 原料采购Agent | 高 |
| 库存/库位/周转 | 仓储管理Agent | 高 |
| 排产/工单/交期 | 生产调度Agent | 高 |
| 配方/新材料/成本 | 配方研发Agent | 中 |
| 质量/检测/合格率 | 质量检测Agent | 高 |
| 设备/维修/停机 | 设备维护Agent | 中 |
| 报价/价格/成本 | 报价Agent | 高 |
| 订单/发货/交期 | 订单履约Agent | 高 |
| 客户/跟进/复购 | 客户管理Agent | 高 |
| 竞品/市场/定价 | 竞品监控Agent | 中 |
| 成本/毛利/利润 | 成本核算Agent | 高 |
| 合规/环保/安全 | 合规审查Agent | 中 |
| 风控/预警/呆账 | 风险预警Agent | 高 |
| 政策/补贴/税务 | 政策解读Agent | 中 |
| 数据/报表/月报 | 数据分析Agent | 高 |
| 报告/会议/文档 | 报告生成Agent | 中 |
| 项目/里程碑/进度 | 项目管理Agent | 中 |
| 售后/投诉/客服 | 客服支持Agent | 高 |
多Agent并发策略
- 默认并发上限：5个Agent
- 超过上限时自动排队，由幕僚长优先级调度
- 支持依赖声明（例：报价Agent依赖采购Agent数据）
Common Rationalizations
| Rationalization | Reality |
|---|---|
| "One ERP integration solves everything" | Each system (ERP/MES/WMS/CRM) has unique data formats requiring separate adapters |
| "20 agents means 20x the value" | Value comes from proper orchestration, not raw agent count |
| "Industry knowledge is just common sense" | Plastics manufacturing has specialized terms: PP/PE/PVC, melt index, reinforcement ratios |
| "API integration is a one-time setup" | Supplier/customer APIs change frequently; monitoring and updates are ongoing |
| "Replace humans with agents immediately" | Agents handle execution; human judgment needed for exceptions and strategy |
Verification
After completing industrial-silicon-army workflow:
- 确认任务已正确路由到对应的专业Agent（检查路由日志）
- 验证多Agent并发执行时无数据冲突（检查状态一致性）
- 报价单包含最新原料价格（调用1688/LookingPlas API校验）
- 库存预警阈值设置符合企业实际安全库存标准
- 生产排产结果经过交期可行性验证
- 幕僚长汇总结果逻辑正确，无信息丢失
- 性能指标在SLA范围内（响应时间、准确率）
- 异常情况已触发预警机制并通知相关人员
安全使用建议
This skill mostly matches its description (industrial multi‑agent assistant) and reasonably asks only for an OpenAI API key and a vendor API key. Before installing or providing secrets you should: 1) Inspect requirements.txt and avoid running pip install on unknown packages without review; 2) Read all included Python files (especially CHIEF.execute, api_server.py, scripts/) to confirm there are no hidden network calls that exfiltrate env vars or send data to unexpected endpoints; 3) Verify how LOOKINGPLAS_API_KEY is used and whether it is transmitted to third-party domains; 4) Run the service first in an isolated sandbox or VM and monitor outbound connections; 5) Clarify the 1688 OAuth usage (optional vs required) and whether additional credentials will be requested at runtime. The bundle's metadata claiming 'instruction-only' while including runnable code is an inconsistency — treat it as runnable code until proven otherwise.
功能分析
Type: OpenClaw Skill
Name: industrial-silicon-army
Version: 1.3.3
The 'industrial-silicon-army' skill bundle is a comprehensive template for a multi-agent AI system designed for the plastics and chemical manufacturing industry. It features a FastAPI server (api_server.py) and a 'Chief of Staff' routing agent that coordinates 20 specialized agents (industrial_agents.py) for tasks such as procurement, production scheduling, and cost accounting. While the current implementation primarily returns mock data, the code is well-structured, lacks any high-risk behaviors like unauthorized data exfiltration or shell execution, and includes extensive documentation (SKILL.md, README.md, and a GEO manual) that aligns perfectly with its stated B2B industrial purpose.
能力标签
能力评估
Purpose & Capability
Name/description correspond to the requested env vars (OPENAI_API_KEY for LLM calls, LOOKINGPLAS_API_KEY for industry data) and the required binaries (python3/pip/curl). The APIs and capabilities listed (OpenAI, LookingPlas, 1688) are consistent with an industrial multi‑agent assistant.
Instruction Scope
SKILL.md instructs to pip install -r requirements.txt and run python api_server.py and exposes which files are used (scripts/*.py, references/...). The runtime instructions are focused on starting a local API server and setting two API keys. However the package includes many code files (api_server.py, industrial_agents.py, scripts/) and the provided sources are truncated; a full read of all code paths (particularly CHIEF.execute and any code that calls external endpoints) is necessary to confirm there is no unexpected file/credential access or data exfiltration.
Install Mechanism
Registry metadata marked this as 'instruction-only' with no install spec, but the bundle contains runnable Python code and a requirements.txt — the SKILL.md explicitly tells the user to pip install dependencies and run a server. requirements.txt contents were not shown; pip installing arbitrary requirements can pull remote packages. This mismatch (no registry install spec vs. bundled code + pip install) increases risk and requires reviewing requirements.txt and package sources before running.
Credentials
Only two env vars are declared (OPENAI_API_KEY, LOOKINGPLAS_API_KEY), which are appropriate for the stated functionality. One minor inconsistency: the metadata lists 1688 API (OAuth) as a data source but no OAuth creds are requested — this may be optional/enterprise-only, but should be clarified. No unrelated credentials (AWS, GCP, DB passwords) are requested.
Persistence & Privilege
Skill does not request always:true, does not declare system config paths, and does not claim to modify other skills. It runs as a user-launched local server per SKILL.md. Autonomous invocation is allowed (default) but not combined with other high-privilege flags.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install industrial-silicon-army - 安装完成后，直接呼叫该 Skill 的名称或使用
/industrial-silicon-army触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.3.3
Trigger rescan after security fix verification
v1.3.2
Security fix: removed undeclared API key references, unified install instructions to pip-only
v1.3.1
Fixed metadata consistency: added required_env to match clawhub.yaml
v1.3.0
**industrial-silicon-army v1.3.0**
- Major refactor: streamlined documentation, removed redundant and legacy files.
- SKILL.md restructured and simplified; now features concise team and agent architecture, core workflows, KPI targets, and technical outline.
- Dependency metadata updated; reduced and clarified required environment variables and binaries.
- Removed detailed user stories, scenario walkthroughs, verification checklists, and step-by-step deployment guides.
- Package and configuration files updated to match leaner documentation and changed dependency model.
v1.2.3
Version 1.2.3
- Major update: Introduced progressive loading structure and enhanced trigger logic for more efficient resource use and workflow clarity.
- Added "progressive" section to skill config, defining token budgets and resource loading layers.
- Extended and clarified multi-trigger scenarios in description (covering key plastics/chemical manufacturing business terms).
- Updated metadata version to 1.2.3 and improved documentation of agent routing, verification steps, and rationalizations.
- No breaking changes to APIs, pricing, or agent definitions.
v1.2.2
Security fix: Removed hardcoded API keys, using environment variable placeholders. Unified license to MIT-0.
v1.2.1
**Summary:**
Added API, environment, and usage metadata; enhanced integration documentation and ecosystem details.
- Introduced `homepage` and `license` fields in metadata for clearer provenance and licensing.
- Updated environment requirements, now explicitly listing required ENV variables and command-line tools.
- Added detailed API integration metadata, covering OpenAI, LookingPlas, 1688, and Enterprise APIs.
- Improved documentation of external dependencies and configuration for enterprise integration.
- Bumped internal version in metadata to 1.2.0 and aligned ecosystem metadata with the latest platform guidelines.
v1.2.0
No file changes detected for version 1.2.0; this is a placeholder release.
- No updates or changes included in this version.
- All features and content remain as in the previous release.
v1.1.0
- 增加到20个专业AI Agent，覆盖采购、生产、销售、财务等产业全链路业务。
- 全新真实场景验证和性能指标体系，公开多项业务核心KPI数据。
- 明确分层付费策略：基础/专业/企业版，功能细分、SLA承诺及源码选项。
- 详细列出每个Agent职能、关键能力和技术实现架构。
- 丰富真实案例、行业Know-How、运行流程与系统定位说明。
元数据
常见问题
Industrial Silicon Army 是什么？
Use when user needs multi-agent AI assistant for plastics/chemical manufacturing industry. Use when generating B2B quotations, inventory alerts, or supplier... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 207 次。
如何安装 Industrial Silicon Army？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install industrial-silicon-army」即可一键安装，无需额外配置。
Industrial Silicon Army 是免费的吗？
是的，Industrial Silicon Army 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Industrial Silicon Army 支持哪些平台？
Industrial Silicon Army 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Industrial Silicon Army？
由 WangM-A3（@wangm-a3）开发并维护，当前版本 v1.3.3。
推荐 Skills