Introducing the Indigo Claude skill

[siclark]

I have been working with LLMs developing software for the last 9 months which has been a fast learning curve. From Copilot in VSCode to ChatGPT on the web and then Claude CLI across many different LLMs.
I have watched the models improve and the ways you use them leap forwards. From simple prompting to structured workflows, then MCPs for connectivity and data as well as planning and task execution. More recently agents and sub agents, in the background and most recently in parallel.

And now Claude has skills. Anthropic says it best here. https://platform.claude.com/docs/en/age ... s/overview.
The key part is...

Why use Skills
Skills are reusable, filesystem-based resources that provide Claude with domain-specific expertise: workflows, context, and best practices that transform general-purpose agents into specialists. Unlike prompts (conversation-level instructions for one-off tasks), Skills load on-demand and eliminate the need to repeatedly provide the same guidance across multiple conversations.

Key benefits:

Specialize Claude: Tailor capabilities for domain-specific tasks
Reduce repetition: Create once, use automatically
Compose capabilities: Combine Skills to build complex workflows

So.. talking to autolog today who is also starting to make use of Claude, and I know many others are, I thought we needed a central store of best practise, sdk knowledge and examples for all who want to use claude to develop plugins so I have created a Claude skill for Indigo.

https://github.com/simons-plugins/indigo-claude-skill

Clone to your vscode project and follow the examples. Feel free to amend and create pull requests to update. There are many more with better knowledge of the SDK than me.

Looking forward to the feedback and seeing what people use this to build.

Simon.

[siclark]

WOW WOW WOW
I just created a new folder in vscode, put in an md file of the Netro irrigation API docs and said
"/indigo build me a sprinkler plugin which uses the Netro API documented in this folder. "

15 mins later I had a working plugin other than 1 small bugs.. a ? not & in the API call which was quickly fixed, and the skill docs are already updated to reflect this.

Clearly I would want to add some extra features, but it read the API, put the detail into the sprinkler device, showed me all the zones with names etc..

Very very exciting.

[CliveS]

Hi Simon,

Thanks for the skill

I’ve been through the whole ChatGPT phase too and eventually got fed up with it going round in circles and disappearing down rabbit holes — sometimes even losing code it had written five minutes earlier.

About three months ago I switched to Claude and honestly, what a difference. The Pro level is well worth paying for.

Unlike you, I’ve only got basic Python knowledge, but I don’t really need much more — I just ask Claude. The joy when I “wrote” my first plugin was huge. It also helped me understand what was actually needed and what all the XML files were for.

When Claude released Claude Code I was completely blown away, and now that I’ve installed your Indigo skill (with a bit of help from Claude), it feels like things can only get better.

[siclark]

I have no python skills, well some beginner knowledge, but a lot of perseverance and hyper focus resulting in a lot of time invested in developing with Claude. (And a lot of wasted hours using copilot)

This skill is work in progress (and welcome input from all). It needs more knowledge added and examples to make it even better.

But it should help turn the excellent documentation into something Claude can use as required without clogging context unnecessarily.

[siclark]

New version on GitHub.

Release Notes

Documentation
- Restructured documentation with improved task-oriented navigation and use-case-based quick lookups
- Added comprehensive modular Indigo Object Model reference documentation
Introduced quick reference sections for common tasks and command namespaces
- Enhanced documentation for devices, events, plugin preferences, and subscriptions
- Updated all documentation to support Indigo 2025.1+

[siclark]

New release on Github.

Summary

Updated command-namespaces.md with new Indigo 2025.1 schedule and action group methods: duplicate(), execute(), getDependencies(), moveToFolder(), removeDelayedActions(), displayInRemoteUI()
Expanded api-patterns.md with practical examples for schedule and action group management
Added open-source-contributing.md with the IndigoDomotics GitHub contribution workflow
Updated skill.md with routing and file references for new content

Thanks to Dave for pointing this out.

[durosity]

This looks super cool - I'll try it out as soon as I can afford to renew my Claude sub

[siclark]

If you have a decent sized apple silicon Mac you could try LM studio with a local LLM running... most will use same skills framework...

[Coolcaper]

Just wanted to add my thanks for the skill. Managed to refine and debug 2 plugins I was working on for a while using this. Really takes things to the next level!

[siclark]

Great. I can’t think not not using mg this now. Also with one of the MCP servers depending on your hardware and desire to use open AI to allow control of devices and restart plugins once your agent deploys them really takes things up a notch.