Fix context7.json schema error, validate it in CI

[ProxiBlue]

Follow-up to #104. The context7.json merged there doesn't pass Context7's schema, so none of its settings took effect. My mistake: I checked it was valid JSON, but never validated it against the schema.

What went wrong

The description I wrote was 361 characters. Context7's schema allows at most 200. That is the only validation error in the file.

Context7 re-parsed the repo on 2026-10-05, 12 minutes after #104 merged, so the refresh workflow works. But the settings shown on the library page for /mage-os/devdocs are still all defaults:

• the generated description, not ours
• the default excludeFiles list, without our 13 unlinked pages
• no rules at all

So the unlinked duplicate pages are still indexed in place of the published *_xml pages. For example, cron_groups.md supplies 9 snippets and cron_groups_xml.md supplies none. The code-block language tags from #104 did land, because they live in the markdown, not in the config.

Changes

1. context7.json: description shortened to 181 characters. No other change. The file now validates.
2. New .github/workflows/context7-validate.yml: a CI check that validates context7.json against the Context7 schema with check-jsonschema. Context7 gives no error when the file is invalid, so this makes a schema error fail the PR instead of going unnoticed. It runs only when context7.json or the workflow itself changes.

Tested:

• Run locally against the file merged in Add context7.json, tag code fences, refresh Context7 on push #104, the check fails with $.description: ... is too long.
• Against the fixed file it passes, both locally and in CI on the fork.

After merge

The refresh workflow will trigger a re-parse. I'll re-check the index afterwards, to confirm the unlinked pages drop out and the rules are served.

🤖 Generated with Claude Code