[Slice 4] new-blog-post end-to-end (multimode)

[wab]

Parent

#21

What to build

Add the blog post content type end-to-end. Uses the team cross-ref from #24 to validate author. This is the highest-volume content type at Ocobo, so adoption of the publishing flow hinges on this slice.

Concretely:

• Create scripts/schemas/blog-post.schema.js — Zod schema with title, author (team slug), description, date, image, exerpt? (encode the typo as-is, with a code comment pointing to the future fix issue), read?, podcastId?, tags?[].
• Extend scripts/cross-ref-resolver.js to validate author exists in the team registry.
• Tests.
• Extend scripts/validate-content.js to validate blog/fr/ (and blog/en/ if non-empty).
• Create .claude/skills/new-blog-post/SKILL.md — multimode. Interview: slug-first conflict check, author picker (active team members only), autocomplete tags from existing values with warn-but-allow for new ones, optional fields batched, preview before write, cover image via drag&drop into assets/posts/<slug>/.
• Create docs/notion-templates/blog-post.md — property mapping + NotionAI prompt for the Blog Posts DB.

Acceptance criteria

• pnpm validate covers blog/fr/ and passes on current data (or reports specific violations).
• Vitest tests for blog-post.schema.js and the extended cross-ref-resolver.js pass.
• /new-blog-post <notion-url> imports a Notion page, validates the author against the team registry, and refuses on a broken reference with a clear message.
• /new-blog-post (no arg) runs the interview; the author picker lists only active team members.
• Autocomplete proposes existing tags by frequency; a brand-new tag is accepted after explicit confirmation.
• The frontmatter is written with exerpt: (not excerpt:) and the schema comment references the typo cleanup issue.

Blocked by

• [Slice 3] new-team-member end-to-end (multimode) #24