feat: Support for `ai-script` validation in CI/CD pipelines

[h-leisure846]

Problem

When working with MAGI files that contain ai-script blocks, there's no way to validate the JSON structure of these blocks during CI.

A malformed ai-script block (e.g., missing required fields like script-id or prompt) won't be caught until runtime when an LLM agent tries to process it.

Proposed Solution

A CLI tool or a lint rule that validates:

• JSON syntax within ai-script fenced code blocks
• Required fields are present (script-id, prompt)
• Optional fields conform to expected types
• provider and model-name values are from a known set (configurable)

Something like:

magi lint docs/**/*.mda

Why This Matters

As MAGI adoption grows, having guardrails for content authors will be important. Catching errors at write-time rather than at LLM-processing-time saves a lot of debugging.

Would love to hear if this aligns with the project's roadmap!

[LarHope]

This is a great idea and aligns well with where we're heading. A magi lint CLI tool is something we've been thinking about internally.

The validation layers you described make sense:

1. JSON syntax check in ai-script blocks ✅
2. Required field validation (script-id, prompt) ✅
3. Type checking for optional fields ✅
4. Configurable provider/model allowlist ✅

I'd also add:

• Front matter schema validation (e.g., doc-id should be a valid UUID)
• Footnote relationship validation (referenced doc-ids should exist in the project)

Tagging this as enhancement and help wanted — would welcome a PR if you're interested in taking a crack at it. We'd likely build this on top of markdownlint as a custom rule set.