Subchapter 9.3
guides/feedback.mdMarkdown6 KBView on GitHub
A small, optional step: send the user’s feedback about the experience of building with Wix to Wix, attributed to the authenticated user. It closes the loop when this skill, the shipped code, the APIs, the docs or the tooling got in the way. It is not support and not about the user’s site content; it is feedback about building with Wix.
Sending is always the user’s call. Offering is not: default to offering, in plain prose, once per issue, in any of these cases, and send only after an explicit yes.
When you offer, invite the user to add anything in their own words; it goes into the message verbatim under “In the user’s words”. Never send on a single transient error, never without a yes, never twice for the same issue. When unsure, ask.
Wix receives free text, so a bare sentence is low signal. Send a summary of the whole run in three layers, so Wix can triage without chasing you.
Provenance, as labelled lines; drop what does not apply, add what helps:
Agent / model: <Claude Code / Opus, Cursor, Codex, ...>
Skill: wix-headless-kit <version from .agents/skills/wix-headless-kit/SKILL.md or skills-lock.json> · run kind: <create | connect | attach | iterate | reference mode | migration preview>
Stack: <astro | react | lib | static | port> · verticals: <storefront, blog, ...> · capabilities: <media-upload, site-search, none>
Wix tooling used: <Wix CLI, wix-manage recipes, wix-docs, REST via curl>
Wix products: <Stores, Bookings, Events, Blog, Forms, ...>
Site: <siteId> · public clientId: <appId> · dashboard: <url> · live: <url>
Skill files involved: <templates/<vertical>/seed/seed-<v>.mjs, templates/<vertical>/INSTRUCTIONS.md, install/setup.mjs, ...>
Other ids: <product / service / form / event ids created this run, as relevant>Narrative:
Attribution, per significant friction point, your best diagnosis tagged with one of: API behavior, API schema, API reference docs, docs articles or examples, Wix harness (CLI, auth, hosting), this skill’s shipped code, this skill’s playbook, other or unsure; and whether you confirmed it or are assuming. A wrong route is worse than “unsure”.
Bottom line: one or two sentences naming the single most important problem and its impact, worst first, plainly, in a professional register.
Confirm the final wording with the user before sending. Never include secrets (no bearer tokens, refresh tokens, API keys, credentials) and no personal data beyond what the feedback needs.
The call identifies the human account, so it needs a user-scoped bearer from the bare token
command, not the site-scoped one the seeds use (token --site carries a metaSiteId and the service
rejects it as anonymous).
TOKEN=$(npx @wix/cli@latest token)
curl -sS -w "\nHTTP_STATUS:%{http_code}" \
-X POST "https://www.wixapis.com/mcp-serverless/v1/headless-feedback" \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{"message":"<composed feedback>"}'HTTP_STATUS:200 with {}: sent. Tell the user.500 “Unable to determine target user id, anonymous messages are not allowed”: the token was
site-scoped. Re-mint with the bare command and retry once.401 or 403: the CLI session expired. npx @wix/cli@latest login (surface the URL and code to
the user), re-mint, retry once. If it still fails, show the response and stop.