adapto-publish

adapto:publish

The terminal step of the draft-first loop: every other skill writes draft content; this one takes the reviewed drafts live (draft → published), or archives them back (published → archived) as a walk-back. It's stateless — it discovers what's publishable rather than remembering a session.

When to use

• "Publish", "publish my drafts", "take it live", "go live", "unpublish/archive this" — typically after reviewing the seeded/translated drafts on the dev server.

When not to use

• Creating or editing content → the content pipeline (content-create → content-upload) or adapto:translate.
• Just checking the environment → adapto:doctor.

Inputs

• Mode: publish (default) or archive (the inverse / walk-back).
• Optional filters: type (Articles / Pages / collection items), language, or a specific collection / item set.
• Collections enumerated from .adapto/schema.json (if present) or adapto collections list --json.

Outputs

• The selected Articles / Pages / collection items moved draft → published (or published → archived).
• The ledger updated (acted-on pieces → published) and .adapto/calendar.md refreshed.
• A report of acted-on + skipped, with ids.
• Next step: the cycle is complete — suggest the next content cycle (adapto:content-research) or adapto:translate to localize what just went live.

Preconditions

• Preflight with the adapto:doctor checks.
• Hard-block on an authenticated CLI (adapto auth me) and a selected tenant — this skill writes. Confirm the working tenant first; never assume the active one.
• adapto CLI >= 0.1.3.

Plan phase

1. Discover candidates in the relevant status (publish → draft; archive → published):

   adapto articles list --status draft --json [--language <lang>]
   adapto pages list --status draft --json [--language <lang>]
   adapto collections list --json                                  # enumerate collections (or read .adapto/schema.json)
   adapto collections items list <collection_id> --status draft --json
   

2. Present them as type · title · slug · language · status.
3. Let the user select — all or a subset (by type, language, or specific items). "Publish everything I just seeded" = select all; targeted = pick a few.
4. Print a machine-parseable plan: the exact set to act on (count, per type, titles/slugs, languages) and the transition (draft → published or published → archived). List items already in the target state as skipped. Wait for an explicit approve. No cost/token figures. Nothing to act on → say so and stop.