Because of its clean design and comprehensive documentation, LuciadCPillar is well-suited for AI-assisted development. LuciadCPillar also ships with an AI skill that streamlines AI access to its development resources, and helps AI coding assistants understand the library and produce correct LuciadCPillar code.

This article describes the skill content, where to find it in the release and how to enable it for the most common AI coding tools. For general guidance on using AI coding assistants, see General AI guidelines and recommendations.

What does the LuciadCPillar AI skill offer?

An AI skill is a small bundle of Markdown files that an AI coding assistant loads as reference material. The LuciadCPillar skill:

Contents of the skill bundle

The skill bundle ships with the documentation set of every LuciadCPillar release, under:

<release-root>/docs/ai/luciadcpillar/
├── SKILL.md                  -- main skill entry point, loaded automatically
├── concepts.md               -- core concepts, such as Map, Layer, Model, and Controller
├── documentation.md          -- how to navigate the offline and online docs
├── samples-and-api.md        -- where to find samples and API reference per platform
└── getting-started.md        -- minimal map application per language

The bundle is plain Markdown, with no installation and no dependencies. The main SKILL.md is the entry point. The other files are loaded on demand when the assistant needs that specific topic.

Enabling the skill in your AI coding tool

Most AI coding assistants discover skills automatically from a project-local directory or through a configuration file. Copy the docs/ai/luciadcpillar/ directory from the release into the location your tool expects, as described below.

The exact directory and discovery mechanism varies between tools and could change. Check the documentation of your tool if the instructions below are out of date.

Claude Code

Claude Code reads skills from a .claude/skills/ folder at the root of your project. Copy the bundle there: 

<your-project>/.claude/skills/luciadcpillar/

The skill is discovered automatically in the next session.

OpenCode

OpenCode discovers skills from .opencode/skills/ at the root of your project. Copy the bundle there: 

<your-project>/.opencode/skills/luciadcpillar/

OpenCode also discovers skills from .claude/skills/ for compatibility with Claude Code. If you have already placed the bundle there for Claude Code, OpenCode picks it up without a second copy.

Cursor

Cursor reads project rules from .cursor/rules/. The simplest setup is to copy the bundle to .cursor/rules/luciadcpillar/ and add a one-line pointer rule that references it. Refer to Cursor’s documentation on Project Rules for the rule-file syntax appropriate to your Cursor version.

GitHub Copilot

GitHub Copilot reads repository-wide custom instructions from a .github/copilot-instructions.md file at the root of your project. Copy the bundle into your project and add a one-line pointer that references it to the copilot-instructions.md file. For example: 

<your-project>/.github/copilot-instructions.md   -- pointer to the bundle
<your-project>/.github/luciadcpillar/            -- the copied bundle

Copilot loads .github/copilot-instructions.md automatically and can then read the bundle files as it needs them. Refer to GitHub’s documentation on repository custom instructions for the details, including the path-specific .github/instructions/*.instructions.md variant.

Other tools

Any AI coding tool that supports the loading of external Markdown context can use the bundle. These are some generic options that work with most tools:

  • Place the bundle in a project-local directory of your choice and tell your tool to include it in the working context, for example through an AGENTS.md, CONVENTIONS.md, or similar configuration file that your tool reads.

  • Paste the contents of SKILL.md into the system or instructions prompt of the assistant. The assistant can then ask for the other files as needed.

What the assistant needs from you

To give you relevant answers, the assistant needs two pieces of information:

  • Your target language — C++, C# or Android (Java/Kotlin). The assistant uses this to choose the right code snippets, samples, and API reference pages, and to write new code in the matching language.

  • The release root — the path on your machine where you extracted the LuciadCPillar release, if any. The assistant uses this to read documentation, samples, and API reference directly from disk, and derives all sub-paths under the root by itself.

The assistant tries to infer both from the files you are working on. If it cannot, it will ask. You can also tell it explicitly at any time: 

I am working on a C# WPF application.
My LuciadCPillar release is at /path/to/LuciadCPillar_<version>/.

If no release is available locally, the assistant falls back to the online documentation:

The offline documentation in your release is preferred when available, because it matches your installed version exactly.

General AI guidelines and recommendations

Developing with a coding agent

Follow these tips for a productive development experience.

Make a plan

For any application beyond a quick demo, start from a development plan:

  • Ask the agent for a detailed plan in a .md Markdown file, and iterate on that plan with the agent until each aspect of the plan is clear.

  • Break the plan down into smaller steps and let the agent implement them incrementally.

Add more skills

Aside from the LuciadCPillar skill bundle content, your skills files can also hold other project information, such as coding style guidelines and project conventions.

Connect the coding agent to your development environment

Consider connecting the agent to your IDE so that the agent can interact with it, and automatically validate code, run tests, and fix issues. Some IDEs offer a Model Context Protocol (MCP) server for this purpose.

AI tooling options

To support application development, you can set up an AI assistant or a coding agent. You can access these as command line interface (CLI) tools, dedicated applications and browsers, or application plugins. If you install an AI extension or plugin in your preferred development environment (IDE), you get AI support from within IntelliJ or Visual Studio, for example.

AI assistants

Popular plugin options include:

  • GitHub Copilot Business or Enterprise, an AI coding assistant from GitHub and OpenAI. It provides access to a broad set of Large Language Models (LLMs).

  • Claude, which focuses entirely on Claude models, and integrates with most IDEs.

AI coding agents

If you want your AI tools to have more workflow autonomy, you can decide to use a dedicated AI coding agent next to your IDE assistant plugin. An AI coding agent can run independently, make decisions, and execute actions autonomously.

Coding agents can typically integrate with an in-house LLM as well as external LLMs. This flexibility allows you to choose a model suitable to the complexity of each task. For design and planning, you select a more advanced model for example, while a lighter model may be sufficient for implementation.

These are some of your options:

  • OpenCode, an open-source agent without ties to a specific LLM. You can choose to use its free in-house model, or you can provide credentials to other LLMs.

  • Cursor, which can integrate with its own models as well as other LLMs.

  • Claude Code, relying on in-house models.