agentsclimarketplace

Adapto seo wire

Skill adaptocms/adapto-cms-agent-skills/plugin/skills/adapto-seo-wire

A pack of skills for AI coding agents (Claude Code, Cursor) that lets them operate Adapto CMS end-to-end: scaffold projects, design schemas, seed content, translate, run SEO, audit content, roll back.

Install
npx -y skills add adaptocms/adapto-cms-agent-skills --skill adapto-seo-wire

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Make stored SEO metadata actually render — a one-time, consent-gated, framework-aware setup that wires a head-render layer (title, meta, OG/Twitter, JSON-LD) reading the _adapto_seo collection, plus generates llms.txt / llms-full.txt from your content inventory. Edits your app templates (never the read-client), only after you approve the exact changes.

SKILL.md

6.3 KB, as published. Nobody here has run it

adapto:seo-wire

adapto:content-upload stores SEO metadata in _adapto_seo, but the starters render only <title>. This skill makes it render: a one-time, per-project setup that installs a head-render layer reading _adapto_seo (emitting <title>, meta description, OG/Twitter tags, and JSON-LD) and generates llms.txt / llms-full.txt from your inventory. It is consent-gated and only edits your own app templates — never the read-client (forbidden-actions.md; see seo-standards.md for what's rendered).

When to use

  • "Make my SEO/meta render", "wire up the head tags", "generate llms.txt", "the OG tags aren't showing".
  • Once per project, after adapto:content-upload has written _adapto_seo items (or alongside the first upload).

When not to use

  • Storing metadata (that's adapto:content-upload, into _adapto_seo).
  • Writing or uploading content → the content pipeline.
  • Just checking the environment → adapto:doctor.

Inputs

  • The scaffolded frontend in the cwd, and its framework (Next / Astro / SvelteKit — detected as adapto:doctor does).
  • The _adapto_seo collection (read at render time, by target_slug).
  • inventory.md — the content list llms.txt is generated from.

Outputs

  • A head-render layer in the app's own templates (per framework: Next generateMetadata / Astro <head> in the layout / SvelteKit <svelte:head>) that reads _adapto_seo and emits title/description/OG/Twitter/JSON-LD.
  • Generated llms.txt + llms-full.txt at the site root, from inventory.md.
  • If you decline the edits: the exact per-framework snippets written to .adapto/seo-render/<framework>/ for you to paste, plus instructions. Nothing is changed without consent.
  • Next step: restart the dev server and check a page's <head> (and /llms.txt); re-run after big inventory changes to refresh llms.txt.

Wiring (consent-gated)

  1. Detect the framework and locate the layout/head entry point + the static/public dir.
  2. Inform + show the exact files/diffs you'd add/change (the head component + the llms.txt generator/route), noting it edits app templates, not the read-client. Ask as a pickable question: Yes, wire it / Just give me the snippets (plus free-form).
  3. On consent: write the head-render layer (reads _adapto_seo by target_slug; falls back to the content title when no metadata exists) and generate llms.txt/llms-full.txt from inventory.md. Idempotent — re-running updates rather than duplicating. Per framework the real work differs — the scaffolds ship dynamic detail routes with little or no metadata:
    • Astro: add meta/OG/JSON-LD to the single head component (src/layouts/Layout.astro).
    • Next (App Router): dynamic routes (app/**/[slug]/page.tsx, [collection_slug]/[item_slug]) export no generateMetadata — add an async generateMetadata(props) to each (await params), and inject JSON-LD as a <script type="application/ld+json"> in the component tree.
    • SvelteKit: each detail route sets only <title> in <svelte:head> — thread SEO through the route's +page.server.ts load → data prop → <svelte:head> (env is server-private, so the fetch stays in .server.ts).
    • All three ship src/lib/reserved.ts (hides _-prefixed collections + shadow slugs articles/collections/micro-copies/pages) — respect it in nav/sitemap/llms.txt.
    • ⚠️ _adapto_seo visibility: content-upload writes those items as draft, and adapto:publish may have published only the content piece, not its SEO item — so a published site's read-client may not see the metadata. Ensure the piece's _adapto_seo item is published too (or that the fetch includes drafts via the key). Verify a real page's rendered <head> after wiring — don't assume it renders.
  4. If declined: write the snippets to .adapto/seo-render/<framework>/ + a short how-to; change nothing else.
  5. Then tell the user to restart the dev server and verify the <head> + /llms.txt. If you started the dev server, leave it running (conventions.md §14).

Preconditions

  • Preflight with the adapto:doctor checks.
  • A scaffolded frontend in the cwd (run adapto:scaffold first) on a supported framework.
  • _adapto_seo should exist (provisioned by adapto:schema-apply); if absent, the layer still installs and simply renders title-only until metadata lands.
  • adapto CLI >= 0.1.3. No auth/tenant needed (no CMS writes).

Errors and recovery

  • Unsupported / undetected framework → stop; emit the generic snippets to .adapto/seo-render/ and explain.
  • Layout/head entry point not found (starter changed) → don't guess-edit; show the user where to add the component and write the snippet.
  • User declines → snippets-only path (above); never edit without consent.
  • llms.txt would expose unpublished/draft URLs → include only published inventory entries; note the skip.

Forbidden actions

  • Never edit or replace the read-client (src/lib/adapto.ts + the published adapto-client-sdk it imports) — app templates only, and only with consent (forbidden-actions.md).
  • Never make host/file changes without explicit consent — inform, show the diff, wait, then write.
  • Never write CMS content (mutates: false).
  • Never put draft-only or secret data into llms.txt / rendered tags.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.