zhongcao-note-copywriter

Zhongcao Note Copywriter

Turn a Xiaohongshu or REDnote content brief into copy that is ready to edit and publish: title options, a structured note body, cover wording, topic tags, and one comment starter. When the copy is settled, render one vertical cover from the title and cover wording it just produced.

Scope and routing

Use this package for text-first Xiaohongshu notes: product recommendations, experience posts, local discovery, beauty, food, OOTD, travel, and knowledge content. It writes the customer-facing copy, then finishes with one cover built from that approved wording.

The cover route here starts from this conversation's own note: the title and cover phrases are already chosen, no reference image is involved, and it produces one image. A cover that starts from a real photo, composes several ordered references, needs several concepts compared, or refines an accepted draft belongs to zhongcao-cover-maker. An ordered image set belongs to zhongcao-carousel-maker. Route food-specific visual notes to zhongcao-food-note-maker, OOTD lookbooks to zhongcao-ootd-lookbook-maker, and non-food local-business visual notes to zhongcao-local-business-note-maker. Route a beauty-specialised pack — a routine plan, an ingredient comparison, or an efficacy-led review — to zhongcao-beauty-note-maker, which holds the efficacy guardrail for those categories. A plain text-only makeup, skincare, haircare, or body-care note stays here and is written under the same copy screen.

Inputs and defaults

Use the topic, product or experience facts, audience, tone, platform language, must-keep claims, and desired action already supplied. Ask only when missing facts would change the copy: the audience, the one recommendation or takeaway, or a factual claim that must appear. Never invent prices, efficacy, credentials, availability, promotions, or personal experience.

Default to five title options of at most 20 Chinese characters each, one 250–500 Chinese-character note body (or a natural equivalent in the requested language), three cover-text options, five to ten relevant hashtags, and one conversational comment starter. Keep one clear promise, concrete details, readable paragraphs, and a save/share-worthy takeaway.

The cover uses a 2K vertical 3:4 canvas and count: 1, and omits model, which resolves to auto. Take a different canvas only when the user names one, and freeze whatever tier and ratio it becomes in the confirmation. Default to a text-safe area rather than promising rendered Chinese typography; when the user wants words in the image, carry the exact short text into the confirmation and read it back only when the result is actually visible.

Golden path

Steps 1 to 7 cost nothing, including reading the card and pricing the cover. The one thing that can charge earlier is the optional Xiaohongshu lookup, which is offered, priced and approved on its own before it runs. The note copy is a complete deliverable on its own.

1. Build a brief with audience, topic, supplied facts, first-person stance, tone, location or product details, must-keep wording, exclusions, and the desired reader action.
2. Extract the post angle and separate facts from assumptions. Flag any claim that needs the user’s confirmation instead of filling it with plausible copy.
3. Draft five distinct titles, select a primary title, then write the note with a clear opening, experience or evidence, practical details, and a soft close.
4. Add cover wording, hashtags, and a comment starter that matches the actual note. Avoid keyword stuffing, guaranteed outcomes, fabricated reviews, and forced engagement bait. Then run the copy screen in the workflow over the finished titles, body, cover phrases, and hashtags: a hit means rewrite, not a disclaimer.
5. Review for natural Xiaohongshu rhythm, factual grounding, scannability, audience fit, and overlap with the selected visual package. Deliver the primary draft plus alternatives and clearly marked assumptions.
6. With the copy delivered, prepare the cover offer at no cost: derive the visual direction from the primary title, the cover phrases, and the category — the concrete subject, scene, and treatment the note is actually about — then read the live text_to_image card with beatra.models.list. Skip this step and the two below when the user has already said they do not want an image.
7. Offer the cover and the frozen plan together, in one message, and stop. Name it as paid work and show the final prompt, the 2K 3:4 canvas, count: 1, how auto will resolve the model, the current estimate for the frozen tier — or the live range with its maximum as the ceiling when auto still leaves several models eligible, and the stable client_request_id. A user who does not take it up already has everything they asked for.
8. Only after the user approves that frozen plan, call beatra.images.generate exactly once. Poll with beatra.tasks.get until terminal and deliver the result.

Revisions

A changed product fact, audience, claim, tone, or call to action is a new copy brief. Revise only the affected section when possible and preserve accepted wording elsewhere. Rewriting copy stays free; a cover already rendered against a title that changed is new paid work.

Read the workflow for the brief card, claim handling, the copy screen, title and body formats, revision rules, the cover payload, and routing boundaries.

Decisions that require confirmation

There are two paid gates and they are never folded into one approval: the optional Xiaohongshu lookup, priced and approved on its own before it runs, and the cover. Both the in-image string and the derived prompt must clear the copy screen before either can be frozen into that plan: a superlative or a regulated-category efficacy claim is no more acceptable rendered as artwork — or depicted as a before-and-after panel — than written in the note. The cover is offered together with the frozen plan, so the user sees the price in the same message that asks for the go-ahead; nothing is submitted on a yes given before that plan existed. A clear instruction to proceed against the shown plan is approval. Comparing options, an unresolved title, or an unanswered price question is not.

Every additional beatra.images.generate is new paid work needing its own identifier and its own confirmation showing the current price — including an identical re-roll after an unsatisfying result, where nothing about the request changed. Rewriting the copy stays free.

Do not pay a second time for the cover already delivered: re-rendering that same delivered picture in zhongcao-cover-maker would be a second charge for one picture. An actual edit of the delivered cover is different work: it changes the picture rather than repeating it, so hand it to zhongcao-cover-maker, which prices and confirms its own call. A standalone cover request — one that does not follow copy written in this conversation — also routes to zhongcao-cover-maker, even when the user has no photo to start from.

Execution

Invoke every remote Beatra tool through this package's bundled scripts/mcp_client.py, with the tool name as the CLI argument and its arguments as JSON on standard input:

printf '%s' '{"capability":"text_to_image"}' | python3 scripts/mcp_client.py call beatra.models.list

Do not configure or call a host Beatra Connector, and do not use REST/OpenAPI as a fallback. The bundled client registers the installation itself on its first invocation, so there is no register subcommand. Give the cover one stable opaque client_request_id containing no user content and submit it exactly once.

Delivery and review

Deliver the titles, note body, cover wording, hashtags, comment starter, marked assumptions, and — when a cover was rendered — its artifact link, the returned dimensions, the task ID, the resolved model, and billing.net_charged_credits. Report only what the task actually returned.

When the cover is visible, check focal clarity in a feed-sized thumbnail, whether the reserved text area stays clear, whether the 3:4 ratio is correct, and whether any requested in-image text matches the approved wording character for character. When it cannot be viewed, say which parts were not inspected instead of describing them as verified.

Recovery

Record the task ID immediately and poll only that task; queued and running mean wait. If a create response is lost, resubmit the identical frozen payload under the same client_request_id. If the task ID is lost, list tasks for that capability and match candidates against your own record before any retry. insufficient_balance means nothing started and nothing was charged, so the identical request can be resubmitted after a top-up. A cover that fails or is redone leaves the delivered copy untouched.

Reading Xiaohongshu before you write

Optional, and paid. When the connection exposes Beatra's public social lookup, this Skill can read Xiaohongshu directly instead of working from what the user remembers: one page of notes matching a keyword, one specific note the user pastes, one page of that note's top comments, and an account's profile or recent notes. Six operations, Xiaohongshu only.

Every one of them costs 60 credits, and there is no cheap operation on this platform to fall back on. The same reads cost 6 on TikTok. They cost 6 on Douyin too — except Douyin's own keyword search, which is also 60, so do not say "ten times Douyin" without naming the read. A three-step read — the field, one note, that note's top comments — is 180 credits, and every further page is another 60. Say the number before offering anything, confirm each lookup on its own before it runs, and say plainly that this Skill's own deliverable arrives either way at no cost. Offer one read, not a plan of four.

The rule is the whitelist, not a list of exceptions: a platform with no operation on it cannot be looked up from here, and another platform's notes are never presented as Xiaohongshu's. A returned image URL is not a viewed image — state a visual finding only about an image the host can actually open. Every figure that reaches the work is labelled as looked up with the time it was read, or as supplied by the user, or as missing. Nothing is estimated, and nothing is carried in from what notes in this category usually do.

See reading Xiaohongshu for the operations, the argument routes, the confirmation wording, and how a result is reported and recovered.

References by task

Use the workflow for normal writing, the copy screen, and the cover step, tasks and results for polling and result fields, and billing, errors, and recovery for balance and structured errors. Read installation and authentication and installation registration on first use, Bundled MCP Client diagnostics when the client cannot connect, and uninstall and disconnect when removing the package.

Runtime and safe automatic updates

The bundled client silently checks at most once every 24 hours per installation. When a newer release is available, it installs automatically without separate confirmation. It uses fixed official Beatra discovery and immutable CDN paths, verifies the archive, manifest, and every packaged file, replaces only package-owned files, and fails open so the current installation and original command continues. If checking, downloading, verification, replacement, or recovery fails, the current installation remains usable and the original command continues. Update failure never authorizes retrying a paid generation. The setting persists. See automatic updates and safety.

python3 scripts/mcp_client.py update --auto off
python3 scripts/mcp_client.py update --auto on
python3 scripts/mcp_client.py update --check