Agent Skills

The following example demonstrates a Skill designed for the N3uron DataOps and IIoT platform. It is intended as a practical starting point that can be adapted, extended, or used as a reference when building Skills tailored to specific business logic, operational workflows, deployment architectures, and module configuration requirements.

This Skill equips an AI Agent with structured knowledge of the N3uron platform, with a strong focus on Tag model creation, configuration, and large-scale deployment workflows. It helps design tag hierarchies, define tag properties, build reusable templates with custom properties, inheritance, and nesting, and generate template definitions or import-ready CSV files from register maps, datasheets, Excel files, PDFs, images, exported tag lists, and other technical sources.

It supports bulk tag creation, template-instance generation, tag import/export workflows, and configuration generation for different N3uron modules and integration protocols. It also covers SQL Client scripting, MQTT custom payloads and parser scripting, REST API Client and REST API Server, the Scripting module, Derived Tags, Historian, WebVision, and MCP Server configuration.

In addition, it supports diagnostics and industrial analysis using logs, historical data, alarm records, module status, tag metadata, and outputs from the N3uron MCP Server.

Skills definition:

A skill is a reusable capability designed for a specific type of task or workflow.

It packages the instructions, context, and any supporting resources needed to perform that task consistently and effectively. Depending on the implementation, a skill may also include reference materials, files, or scripts that support execution.

Rather than being continuously exposed, a skill remains available in the background and is applied when the situation or request matches its intended use.
Overview - Agent Skills

Note:

N3uron SKILL file might receive updates to incorporate new knowledge about the N3uron DataOps & IIoT platform.

Download SKILL file from this link

Upload the SKILL file to Claude Desktop

Upload the SKILL file to Claude Desktop, go to Customize → Skills → Upload a skill.

You can also build your own skills with Claude’s Skill Builder.

Install manually from the SKILL file in Claude Code

Install manually from the SKILL file in Claude Code. Extract downloaded .zip file, and copy the extracted skill folder into your Claude Code skills directory:

mkdir -p ~/.claude/skills
cp -R n3uron-iiot-dataops ~/.claude/skills/

Upload the SKILL file to ChatGPT

Upload the SKILL file to ChatGPT, go to Profile menu→ Skills → Upload from your computer.

Note:

The interface and location of the Skills settings might change over time.

You can also develop your own skills with ChatGPT or enhance and update an existing one.

Install manually from the SKILL file in Codex

Install manually from the SKILL file in Codex. Extract the downloaded .zip file, and copy the extracted skill folder into your Codex skills directory:

mkdir -p ~/.agents/skills
cp -R n3uron-iiot-dataops ~/.agents/skills/

Install manually from the SKILL file in Antigravity

Install manually from the SKILL file in Antigravity. Extract the downloaded .zip file, and copy the extracted skill folder into your Antigravity skills directory:

mkdir -p ~/.gemini/antigravity/skills
cp -R n3uron-iiot-dataops ~/.gemini/antigravity/skills/

Upload the SKILL file to MCPJam Inspector

Upload the SKILL file to MCPJam Inspector. Extract the downloaded .zip file, then navigate to the Skills tab in MCPJam Inspector to discover and manage skills. Click the + button and select the extracted skill folder.

Tip:
Combine N3uron MCP Server Tools with SKILL Knowledge

After uploading the SKILL file, connect the AI agent to your N3uron MCP Server. The MCP Server provides access to N3uron-specific tools and context, while the SKILL file adds domain-specific knowledge, instructions, procedures, and best practices to improve the quality and relevance of the AI agent’s responses.

SKILL file structure

The N3uron IIoT DataOps Skill is distributed as a single ZIP package containing the Skill instructions, N3uron engineering references, deterministic tooling, reusable assets, and its validation tests.

SKILL.md acts as the control plane: it identifies the user's intent and routes the task to the appropriate workflow and supporting resources. Detailed product knowledge remains in references/, while repeatable or fragile operations are implemented as deterministic Python scripts.

n3uron-iiot-dataops/
│
├── SKILL.md
│   └── Main instructions, workflow selection, safety rules and routing
│
├── agents/
│   └── Skill metadata
│
├── references/
│   └── N3uron engineering knowledge
│       • platform and diagnostics
│       • configuration and tag model
│       • protocols and modules
│       • Configuration Orchestrator
│       • security and integrations
│       • SDD and acceptance
│       • WebVision
│       • engineering examples
│
├── scripts/
│   └── Deterministic engineering tools
│       • Configuration Orchestrator
│       • N3uron configuration generators
│       • .n3c utilities
│       • tag/model utilities
│       • SDD tools
│       • WebVision tools
│
├── assets/
│   └── Reusable machine-readable resources
│       • schemas
│       • templates
│       • configuration contracts
│       • WebVision assets
│
└── tests/
    └── Regression and contract tests
        • Skill behavior
        • Configuration Orchestrator
        • configuration generators
        • N3uron utilities
        • WebVision

How the structure is used

The Skill follows a progressive-loading architecture. The AI Agent does not need to load the entire package for every request.

User request
     ↓
  SKILL.md
     ↓
Select workflow
     │
     ├── Reference / troubleshooting
     ├── Engineering change
     ├── Configuration Orchestrator
     ├── Offline project analysis
     ├── Artifact generation
     └── WebVision engineering
     ↓
Load only the required references
     ↓
Run deterministic scripts when appropriate
     ↓
Validate the result

Recommended use with the N3uron IIoT & DataOps platform

Use the N3uron IIoT DataOps Skill as an engineering knowledge and workflow layer on top of the N3uron platform. For best results, provide the AI Agent with a clear objective and the most relevant N3uron context or technical source available for the task.

• For troubleshooting and diagnostics, provide concrete evidence such as the affected tag path, module instance, log file or log entry, alarm, configuration export, or backup. Include the N3uron and module version when the behavior may be version-dependent. This helps the Skill correlate configuration, runtime status, tag quality, timestamps, and logs instead of diagnosing the problem from assumptions.

  • Example: “Investigate why /PLANT01/UTILITIES/COMP001/OUTLET_PRESSURE reports Bad quality. I have attached the relevant module configuration and log file. Identify the likely cause, correlate it with the tag source configuration and logs, and recommend the next checks.”

• For tag, template, or configuration generation, attach the source material whenever possible, such as register maps, CSV/XLSX files, PDFs, exported tags, existing .n3c files, or project requirements. Ask the Skill to validate the source before generating import-ready artifacts and to avoid filling undocumented gaps with inferred values.

  • Example: “Use the attached Modbus register map to generate an N3uron template for this energy meter. Include only registers with a confirmed address, data type, and documented meaning. Preserve the documented units and access mode, validate the resulting source configuration, and generate the final import-ready Template JSON zipped.”

• For live N3uron environments, combine the Skill with the N3uron MCP Server when the AI Agent needs current tag values, quality, timestamps, module status, logs, alarms, historical data, or other runtime context. The Skill provides the N3uron-specific engineering knowledge and workflow, while the MCP Server provides access to the actual state of the node. This is especially useful when a diagnosis requires correlation between multiple runtime sources.

  • Example: “Using the N3uron MCP Server, investigate why the compressor discharge-pressure tag has stopped updating. Check its current value, quality and timestamp, identify the source module, inspect the module status and recent logs, and compare the latest value with Historian data before proposing a cause.”

• For bulk configuration changes in a live N3uron environment, combine the Skill with the N3uron MCP Server and N3uron's native export/import workflows. Use the MCP Server to inspect the current node, gather runtime context, and verify the result, while the Skill handles deterministic transformations of exported Tags CSV files, Template JSON/ZIP files, or other configuration artifacts. This approach is especially useful when hundreds or thousands of tags must be created or modified consistently, because the output can be validated before it is imported rather than applying many individual changes directly to the live node. The Skill includes dedicated workflows for bulk editing exported N3uron Tags CSV files and for generating reusable Template JSON files that can be packaged as ZIP files for import.

  • Example: “I have exported the existing N3uron Tags CSV for CAB01. Create equivalent groups and tags for CAB02 through CAB10. Preserve the original CSV structure and all existing tag properties, update only the equipment path and the corresponding OPC UA NodeId references, and enable Historian only for the measurement tags. Validate the generated CSV before import. After I import it into N3uron, use the N3uron MCP Server to verify representative tags from each new cabinet, including their value, quality, timestamp, source configuration, and Historian behavior.”

• For WebVision projects, provide as much existing project and design context as possible. In addition to the available tag model or exported tag list, you can attach screenshots, mockups, images, SVG files, existing containers.n3c files, or WebVision backups from the current project or from another project that can be used as a design reference. These artifacts help the Skill understand the expected layout, visual language, component composition, bindings, reusable templates, DataSets, Functions, aliases, navigation patterns, and SVG techniques instead of recreating them from a textual description alone. Where possible, also describe who will use the screen, what operational question it should answer, which decisions the operator needs to make, and how states such as alarms, Bad quality, stale data, or expected-zero values should be represented. A supplied containers.n3c can be used as evidence of an existing WebVision structure and as a reference when creating a new project, screen template, reusable widget, visual component, Function, or DataSet. Existing project artifacts should be treated as implementation examples for the relevant WebVision version rather than assumed to be universal configuration contracts.

  • Example: “Create a new WebVision inverter-detail screen using the attached tag export and the existing containers.n3c from another plant as a reference. I have also attached a screenshot of the desired layout and the SVG used for the inverter graphic. Reuse the visual and interaction patterns where appropriate, but bind the new screen only to the supplied tag model. Create the required screen template and reusable components, add any necessary DataSets or Functions, preserve the existing project conventions, and generate the resulting WebVision containers.n3c for validation before deployment.”

Tip:

The Skill produces better results when the request includes concrete N3uron evidence and a measurable engineering objective. For example, “Investigate why this Modbus Client tag has Bad quality using the attached logs and configuration” is more actionable than “Check my Modbus configuration.”

N3uron Skills Prompt example for Document-Based Template Generation

An example prompt that shows how to use N3uron Skills to generate templates from a technical document. It helps the skill read the document as the main reference, extract only the information that is clearly defined, validate the results, and create clean N3uron templates ready to import. It also includes rules to avoid guesses, unclear data, or incorrectly structured outputs.

Use the n3uron-iiot-dataops skill and its template-generator workflow to read the attached SolarEdge SunSpec Modbus PDF first, extract the documented Modbus-over-TCP monitoring map, and generate N3uron Modbus TCP templates as separate JSON templates, each packaged in its own ZIP file.

Goal:
Create independent template outputs for each documented logical element shown in the PDF. Do not merge them into a single template. Generate separate templates for:
- Common
- Inverter model block 101/102/103
- Model 160 Multiple MPPT extension
- Meter 1
- Meter 2
- Meter 3

Source-of-truth rule:
- Read the uploaded PDF first and use it as the only source of truth.

Extraction rules:
- Extract only explicitly documented Modbus-over-TCP monitoring registers.
- Use only confirmed entries: include a register only when all of these are clearly documented in the PDF:
  - address
  - name
  - register/data type
  - size/length
- Preserve the documented SunSpec structure and keep tags grouped exactly by the PDF’s own sections/models.
- Treat Common and the inverter model block 101/102/103 as separate mappings.
- Treat inverter models 101, 102, and 103 as one shared template block, because the PDF documents them as one common register table distinguished by `C_SunSpec_DID`, not as three separate tables.
- Treat Model 160 as a separate template.
- Treat Meter 1, Meter 2, and Meter 3 as separate templates.
- Include documented units and access mode only when explicitly stated.
- Include documented scale-factor registers only as standalone documented tags when present in the PDF.
- Do not apply scaling to any value tags.
- Do not convert scale-factor registers into N3uron scaling blocks.
- Do not infer which scale factor applies to a value unless the PDF states it explicitly enough to implement deterministically.
- Even where the PDF describes the general SunSpec scale-factor concept, do not implement scaling logic in the template; leave scaling out entirely because I will handle scaling manually.
- Do not infer missing fields, undocumented gaps, unsupported points, or unstated mappings.
- Omit any ambiguous, contradictory, malformed, duplicated, or inconsistently documented entries.
- If the PDF contains conflicting addresses, broken rows, table corruption, or section-level inconsistencies, exclude those entries rather than guessing.

History rules:
- Enable history for measurement/value tags.
- Disable history for identification/static metadata tags.
- Disable history for standalone scale-factor tags.

Template split requirements:
- Produce one JSON template per documented logical element/section.
- Produce one ZIP file per JSON template.
- Each ZIP must contain exactly one JSON template file and nothing else.
- Do not bundle multiple JSON templates into a single ZIP.
- Use clear deterministic filenames:
  - common.zip containing common.json
  - inverter_model_101_102_103.zip containing inverter_model_101_102_103.json
  - model_160_mppt.zip containing model_160_mppt.json
  - meter_1.zip containing meter_1.json
  - meter_2.zip containing meter_2.json
  - meter_3.zip containing meter_3.json

TDD requirement before final delivery:
Before generating the final ZIP files, perform an internal test-driven validation pass and reject any bad data.
At minimum validate:
1. Every included tag maps to a register explicitly present in the PDF.
2. Every included tag has confirmed address, type, and size/length.
3. No inferred tags are present.
4. No scaling logic has been implemented in any template.
5. Scale-factor registers, if included, remain standalone documented tags only.
6. Each JSON file is structurally valid for N3uron template import.
7. Each template contains only tags belonging to its own logical section.
8. Each ZIP contains exactly one JSON file.
9. Ambiguous or contradictory rows are excluded.

Output constraints:
- Do not return explanations, summaries, markdown, tables, logs, or commentary.
- Do not return intermediate CSVs or preview data.
- Do not return a combined monolithic template ZIP.
- Return only the final ZIP files, one ZIP per template JSON.

Customizing and extending it

• Add your site topology, tag namespace, and known quirks to SKILL.md so the AI model has permanent context for your deployment.

• Extend the protocol checklists in protocols.md with your equipment-specific failure modes.

• Add sections to report-template.md for your organization's reporting format.

• Document your Custom MCP Tools in the Skill so the model knows when and how to call them — enabling fully bespoke workflows backed by N3uron Scripting.