agentsclimarketplace

Iaiops

Skill industrial-aiops/industrial-aiops/skills/iaiops

Install
npx -y skills add industrial-aiops/industrial-aiops --skill iaiops

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Vendor-neutral, governed industrial/OT data tap + intelligent troubleshooting. Read (and, gated, write) PLCs, controllers, machine tools and IIoT brokers over OPC-UA, Modbus-TCP, Siemens S7comm, Mitsubishi MC, Omron FINS, MTConnect, MQTT/Sparkplug B, Allen-Bradley EtherNet/IP, EtherCAT (pysoem/SOEM), SECS/GEM (semiconductor / display fab equipment over HSMS), PROFINET (DCP discovery), the building edition (BACnet/IP), and Phoenix Contact PLCnext vPLC — plus cross-protocol diagnostics ("no-data" dataflow diagnosis, OPC-UA connection self-diagnosis, subscription health, ISA-18.2 alarm bad-actors, tag/historian health, and the AI downtime root-cause copilot) and analytics (OEE/downtime, asset inventory, OPC-UA HDA, change-of-value). Use when the task names any industrial protocol, a PLC/SCADA/HMI/historian/CNC/RTU/IED, a semiconductor/display fab or SECS/GEM equipment, an electrical substation, an opc.tcp:// or mqtt:// endpoint, OEE/downtime, downtime root-cause, or OT asset inventory. Routes to the iaiops MCP server. Read-first; writes are MOC-gated (high risk, dry-run + double-confirm). Do NOT use for IT/network gear, Kubernetes, hypervisors, or backups — those are separate AIops tools.

SKILL.md

7.0 KB, ~2.3k tokens by cl100k_base, as published. Nobody here has run it

iaiops — ROUTER（识别现场 → 派发 edition skill + MCP profile）

这是路由 skill，本身不含工具表。识别任务的行业/协议 → 转到对应 edition skill （工具清单、workflow、支持版本矩阵都在那边），并用对应 profile 启动 MCP（IAIOPS_MCP=<x> 环境变量，或等价 sugar 脚本 iaiops-mcp-<x>）。现场只暴露 1-2 个协议 + 跨协议脑， 单会话工具 ~15-30，模型不被淹。

路由表（关键词 → edition skill → MCP profile）

任务提到edition skillMCP profile / entrypoint
SECS/GEM · SECS-II · HSMS · 半导体/显示 fab · wafer · panel · MES equipment · SVID/ECID · recipeiaiops-fabIAIOPS_MCP=fab / iaiops-mcp-fab
离散制造产线 · PLC（S7/S7-1200/1500 · 三菱/MELSEC · Allen-Bradley/ControlLogix/CompactLogix · Omron/FINS/SYSMAC CS/CJ/CP/NX）· Modbus · EtherCAT/CoE/SOEM · PROFINET/DCP · MTConnect/CNC 机床 · Sparkplug B/UNSiaiops-factoryIAIOPS_MCP=factory / iaiops-mcp-factory
流程工业（化工/制药/食品饮料）· HART/HART-IP · 变送器/阀门定位器 · 过程仪表 · DCS 旁路读取iaiops-processIAIOPS_MCP=process / iaiops-mcp-process（有 UNS 时 IAIOPS_MCP=process,sparkplug）
BACnet/BACnet-IP · HVAC/暖通 · BMS/楼宇自控 · 厂务/facility · Who-Is · TrendLogiaiops-buildingIAIOPS_MCP=building / iaiops-mcp-building
水处理/水厂 · pH · 浊度/turbidity · 电导率/conductivity · 加药/dosing · 泵站/pump station · 曝气iaiops-waterIAIOPS_MCP=water / iaiops-mcp-water
仓储/物流中心 · 输送线/conveyor · 分拣/sorter · AGV/AMR · WMS/WCS · 托盘/穿梭车 · 物料搬运iaiops-warehouseIAIOPS_MCP=warehouse / iaiops-mcp-warehouse
医院设施/clinical facility · 负压/正压隔离病房 · 医用气体 · 手术室环境 · 医疗 BMSiaiops-clinicalIAIOPS_MCP=clinical / iaiops-mcp-clinical
光伏/PV · 逆变器/inverter · 组串/string · 风电/风机 · 场站 SCADA · 新能源电站iaiops-renewablesIAIOPS_MCP=renewables / iaiops-mcp-renewables
Phoenix Contact PLCnext / vPLC（虚拟化 PLC，走内建 OPC-UA + Modbus-TCP）iaiops-plcnextIAIOPS_MCP=plcnext / iaiops-mcp-plcnext
IO-Link · IO-Link master · IODD · 智能传感器/executor 参数 · pdin · ISDU（主站 JSON/REST 接口，只读）iaiops-factory（楼宇传感器场景 iaiops-building）IAIOPS_MCP=factory / 单协议 IAIOPS_MCP=iolink / iaiops-mcp-iolink
  • 单协议深用（只有一种设备）：直接 IAIOPS_MCP=opcua（或 modbus/s7/mc/fins/eip/mtconnect/ sparkplug/ethercat/secsgem/profinet/bacnet/hart），等价脚本 iaiops-mcp-<协议>。
  • 跨行业只问 OEE/停机根因/资产盘点/数据质量：任一 profile 均可 —— 跨协议脑 （诊断/OEE/资产/合规）默认随 server 暴露，选覆盖现场协议的 profile 即可； 纯脑场景可用 IAIOPS_MCP=brain / iaiops-mcp-brain（只暴露脑，零协议）。
  • 说不清行业但报了协议名：按协议列选含它的最小 profile；都不沾 → iaiops doctor + protocols_supported 先看现场配置了什么。

部署（0.10.0 起：无默认，必须选菜单）

  • 裸启动 iaiops-mcp（不设 IAIOPS_MCP）不再暴露全部工具：打印选择菜单到 stderr 并 exit 2。必须设 IAIOPS_MCP=<selection> 或用预置入口 iaiops-mcp-<name>。 IAIOPS_MCP=menu 显式打印同一菜单（带各 selection 工具数）后退出。
  • IAIOPS_MCP=all 仍可显式使用（power user；>100 工具时启动日志给洪泛警告）。
  • 多进程站点（1 脑 + N 协议）：跑一个 iaiops-mcp-brain（只脑），协议 server 各自 加 IAIOPS_MCP_NO_BRAIN=1（如 IAIOPS_MCP_NO_BRAIN=1 iaiops-mcp-opcua）去掉脑， 避免跨 server 工具重名；protocols_supported（发现工具）在 NO_BRAIN 下仍然保留。

Energy 协议 → 另一个包（不在本 server）

IEC 60870-5-104 / IEC-104、DNP3 / outstation、IEC 61850 / MMS / IED、变电站/ substation、电力/能源/utility SCADA → 用独立的 iaiops-energy （pip install iaiops-energy，然后 iaiops-energy-mcp； github.com/industrial-aiops/industrial-aiops-energy）。它复用本包的 core + 脑； monitor-direction only。不要尝试用本 server 的 profile 覆盖能源协议。

安全不变量（所有 edition 一致，路由前先记住）

  • Read-first：默认只读；先 protocols_supported 看能力、iaiops doctor / <protocol> doctor 验链路，再谈读，最后才谈写。
  • 写 = MOC（Management-of-Change）：The 10 write tools（s7_write_db、 mc_write_words、fins_write_words、mqtt_publish、eip_write_tag、ethercat_write_sdo、 ethercat_set_state、profinet_dcp_set、bacnet_write_property、bas_command）全部 risk=HIGH、默认 dry_run=True、捕获改前值供 undo（bas_command 另对生命安全点一律拒绝）。
  • 审批：真实写入需具名审批人 —— CLI iaiops approve 双重确认后才放行。
  • 未经授权绝不写生产控制系统；AI 结论仅 advisory，引用真实信号出处，宁可 insufficient_evidence 不臆测。
  • 各协议的验证状态（✅ 已自测 / 待核实）见 edition skill 的支持版本矩阵， 标 待核实 的不得当既成事实。

Setup（各 edition 相同）

iaiops init 交互写 ~/.iaiops/config.yaml；凭据进加密存储 （~/.iaiops/secrets.enc，主密码 IAIOPS_MASTER_PASSWORD）。iaiops doctor 逐端点探活。协议参数、模拟器自测、MCP JSON 示例见 README 与各 edition skill。

Gives 0 of the 12 instructions most mcp tooling skills give in ~2.3k tokens

Counted across 638 of the 750 authors here whose files we hold, read 2026-08-06

  • create ten complex read-only evaluation questionsin 71 of 638, across 17 files
  • test servers using MCP Inspectorin 60 of 638, across 18 files
  • provide actionable error messagesin 56 of 638, across 14 files
  • prioritize comprehensive API coverage over specific workflowsin 54 of 638, across 12 files
  • use TypeScript and Streamable HTTP for remote serversin 53 of 638, across 7 files
  • define structured output schemas where possiblein 51 of 638, across 9 files
  • use Zod or Pydantic for input schemasin 48 of 638, across 6 files
  • fetch MCP specification pages with markdown suffixin 46 of 638, across 4 files
  • load framework documentation using WebFetchin 45 of 638, across 3 files
  • verify each evaluation answer independentlyin 45 of 638, across 3 files
  • implement API client with authentication and paginationin 45 of 638, across 3 files
  • Define input schemas with validationin 28 of 638, across 10 files

Said here and by no other author read

  • match industry or protocol to the correct edition skill
  • use single-protocol profile for single device tasks
  • use the brain profile for pure analytics tasks
  • run discovery and diagnostics before any read
  • treat all write operations as management of change
  • require named approver and double confirmation for real writes

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.