Adapto microcopy

adapto:microcopy

Manages a project's micro copy — the short UI strings (nav, buttons, labels, validation, empty/loading states, footer) stored in Adapto as key/value/language entries. Two modes:

• init — seed a curated, project-aware starter set of UI strings with on-brand values (greenfield).
• extract — scan an existing frontend for hardcoded UI strings, create entries, and emit a file:line → key replacement guide (it never rewrites your source).

⚠️ Micro copy has no draft state — entries go live on create. So the plan-approval below is the only safety gate (there's no draft-review net like the content skills have). Review the plan carefully.

When to use

• init: "seed UI strings", "starter microcopy", "set up my UI text".
• extract: "extract microcopy", "pull the hardcoded strings from my site into Adapto".

When not to use

• Translating microcopy into another language → adapto:translate.
• Content (articles / pages / collection items) → adapto:content-seed.
• Just checking the environment → adapto:doctor.

Inputs

• Mode — init or extract. Infer from the request; if ambiguous, ask which.
• Project context for voice (.adapto/project.md cache or adapto collections get-by-slug _adapto_project_config --json); absent → neutral defaults.
• Tenant language (adapto auth orgs --json).
• For extract: a scaffolded frontend in the current directory.
• Key convention: dot-namespaced (nav.home, button.submit, form.email.label) — both modes use it.

Outputs

• Created micro copy entries (key/value/language); a report of created + skipped.
• extract only: a replacement guide at .adapto/microcopy-extract-<date>.md mapping file:line → key, so you can wire the read-client lookups yourself. Source is never modified.
• Next step: suggest adapto:translate to localize the micro copy into another enabled language; for extract, also wiring the read-client lookups from the guide; for init, extract mode to catch any hardcoded strings already in the frontend.

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. extract additionally needs a scaffolded frontend in the cwd.

How each mode gathers its candidates

• init — read project context → auto-propose a curated, dot-namespaced starter set (nav, buttons/CTAs, form labels, validation/errors, empty/loading, footer; tailored to the project type — e.g. e-commerce → cart/checkout, a blog → read-more/subscribe) with on-brand values → take one edit pass. Values follow the applicable prose-standards.md rules (plain language, no filler/jargon — §6; rhythm/paragraph rules don't fit UI strings).
• extract — detect the framework (Next/Astro/SvelteKit, as adapto:doctor does) → scan .astro/.tsx/.jsx/.svelte for user-facing strings (template/JSX text + placeholder, aria-label, title, alt; skip code, imports, className) → propose a dot-namespaced key + value per string. LLM = Sonnet-class.

Plan phase

Print a machine-parseable plan and wait for an explicit approve:

• The keys + values to create, the target language, and which keys already exist (skip via get-by-key).
• extract also shows the source location (file:line) for each proposed key.
• ⚠️ No draft — entries go live on create, so this plan is the review. No cost/token figures.
• Nothing to create → say so and stop.

Apply phase

Loop (no batch); --json on each; dedup via get-by-key:

adapto microcopy get-by-key <key> --json        # skip if it already exists
adapto microcopy create --key <key> --value "<value>" --language <lang> [--tags "<tags>"] --json

• extract only, after the creates: write the replacement guide .adapto/microcopy-extract-<date>.md (file:line → key). Never rewrite source.
• Report created + skipped (with keys). Partial failure (no batch) → report what was created, stop; re-run is safe. Loop cleanly — judge success from each call's --json, end the loop exit 0 on success, so a created batch never surfaces as a red Error: Exit code 1 (§8).
• Then restart the dev server (stop→start) and keep it running so the user sees the new micro copy — never kill it (starters sync content at startup — §14).

Errors and recovery

• extract with no frontend in cwd → stop; it scans a scaffolded project — run it from the project root.
• Key already exists → skip (idempotent); report it, don't overwrite (use microcopy update manually if intended).
• Partial failure mid-loop (no batch) → report what was created, then stop; re-run is safe.
• extract finds nothing (or only false positives) → report; nothing to create.
• Not authenticated / no tenant → stop; offer both auth paths — Log in or Register (conventions.md §11) — then tenant selection.
• Language discovery fails → ask the user for a language code the tenant has enabled; don't guess.

Forbidden actions

• Never write without an approved plan — and since micro copy has no draft, the plan is the only gate.
• Never overwrite an existing key (dedup via get-by-key); skip and report instead.
• Never rewrite source in extract — emit the replacement guide only (don't touch the frontend/read-client).
• Never assume the working tenant — confirm it before any write.
• Plus the global list in forbidden-actions.md — notably: never delete CMS content (archive instead; conventions §9a).