Create Agent Skills for easy Knowledge Base generation

[david-debest]

Goal

Create a GitHub Copilot agent skill that guides users through defining their Knowledge Base interactively and then generates the corresponding code or config.

Behaviour

The skill operates in two phases:

Phase 1 — Grilling:

Start with a natural language description of the user's domain and KB purpose, then drill down into:

• Which KI types are needed (ANSWER/REACT for providing knowledge, ASK/POST for requesting it).
• The graph pattern(s) — what shape of RDF data is exchanged.
• Prefixes and ontology URIs.
• Data source type (Python custom logic, SPARQL endpoint, etc.).
• KB identity (ID, name, description) and KE endpoint URL.

Phase 2 — Generation:

Ask the user which output mode they want:

• Python file — a complete .py file with KnowledgeBase instantiation, KI registrations (decorators for ANSWER/REACT, method calls for ASK/POST), typed BindingModel classes, and handler stubs. Compatible with knowledge-mapper run.
• Config YAML — a config.yaml compatible with knowledge-mapper sparql (or other config-driven subcommands).

Implementation notes

• Skill should be placed in the repository under a skills/ directory and documented.
• Use CONTEXT.md for domain language and API reference during generation.
• Generated code should follow the patterns in examples/.