Proposal: automate CONTRIBUTING skill quality checks (open to contributing additions)

[2rob]

Hi — I'm from ModelBound. We build open-source tooling for agent skills (lint, trust heuristics, token hygiene) and maintain modelbound-mcp on npm. We've been looking at repos like this one where skill quality matters a great deal given the quantity of skills.

Your CONTRIBUTING already lists four automated checks under Review Process as (when implemented):

1. File structure validation
2. Code block language detection
3. Documentation size check
4. Marketplace.json validation

You already have inventory CI (check-inventory.sh); the four above are still mostly manual for contributors.

Idea: land these incrementally in small PRs. We'd be glad to help where our OSS tooling fits, and follow your lead on repo-specific rules (marketplace layout, reference size targets, etc.).

Check	Possible approach
SKILL.md lint (frontmatter, spec, size warnings)	modelbound-mcp in CI — lint mode, no API key
Code block language tags	Small repo script
references/ size targets	Small repo script
marketplace.json validation	Small repo script

A few questions before we open a PR:

1. Should new checks block merges or warn at first?
2. Changed skills on PRs only, or full library on main?
3. Any preference on order (lint first, then the rest)?
4. Should anything run before sync-skills.yml?

If this direction sounds reasonable, we'll follow with a small lint-only CI PR (no secrets, no cloud). Happy to adjust based on your feedback.