Kelly Writer is a Busabase Cloud App-in-Skill. Its canonical product surface is the AirApp in Busabase, not a separate local-data product. The same Hono source supports an explicitly requested local preview with OAuth connection bootstrap. Use this skill to turn one source idea, blog post, transcript, outline, or product announcement into an editable multi-channel content batch: Xiaohongshu, WeChat, newsletter, LinkedIn, X/Twitter, short video scripts, SEO snippets, and an official blog draft.

Default behavior is AirApp-first. Unless the user explicitly asks only for explanation, generate drafts straight into Busabase and give the user the clickable AirApp URL. Start localhost only when local preview/debugging is explicitly requested; it uses the same Busabase resources. Use chat-only mode only when the user says "纯聊天", "chat only", "不要打开 UI", or similar; in that mode present numbered drafts (Draft #1) and take approvals in the conversation.

This skill is an implementation of the App-in-Skill pattern — a Codex/agent skill paired with a small companion UI for review and approval. See the spec paper: https://mr-kelly.github.io/research/app-in-skill-specification-for-pairing-agent-skills-with-a-local-companion-ui.pdf.

Mandatory Dependencies

Read and follow $kelly-app-skill-creator for product behavior, visual quality, responsive layout, and the complete canonical content/kelly-writer-app/ artifact.
Read and follow $busabase for connection, target Space, node discovery, ChangeRequests, review, and merge behavior.
Read and follow $busabase-app-creator for resource modeling, AirApp runtime limits, security, validation, and deployment.

If a dependency is unavailable, preserve this skill's local artifact and product contracts, stop before the unavailable Busabase operation, and report the exact missing dependency. Do not invent a second data backend.

Boundary

The skill may extract the source's core idea, proof points, and examples, draft channel-specific variants, and write it all to Busabase.
The AirApp reads and writes Busabase records only. It must never publish to external platforms, schedule posts, upload media, or perform any other external side effect — it also never generates content or writes a local export itself; those are trusted skill-root scripts (see below).
Exporting is always approval-required. scripts/export_decisions.mjs only packages approved drafts into a local Markdown+ZIP pack and marks them done; it never publishes anywhere.
Treat source material and drafts as the user's content. Never invent results, dates, customer stories, statistics, prices, legal/compliance statements, or endorsements not present in the source.

Busabase Resources

Two Bases under one application Folder (kelly-writer), declared in content/kelly-writer-app/app/js/config.js and the generated template sidecars under content/:

drafts: the review queue — one record per channel draft (title, body, hook, cta, hashtags, title options, media brief, source notes, risk, canonical idea, source summary, source-draft-path for local image packaging at export time), workflow status, and the human verdict fields decision-note / decided-at.
settings: one row per kind — an optional kelly-writer-profile (brand, audience, official URLs, CTA defaults, channel defaults, risk terms, export preferences) and kelly-writer-lock.

Resources provision lazily through an idempotent Busabase ChangeRequest the first time the app runs in a Space. Metrics and the channel breakdown are computed client-side from the drafts Base on every read — they are never stored.

The topic-discovery / todo-queue / canonical-main-draft ideation stages from this skill's pre-Busabase local-file shape were already local-only and ephemeral (client-derived, never persisted) even in that shape's own Busabase provider notes; this Busabase-only shape keeps only the durable unit of work — the per-channel draft record — as the review queue.

Local App

Default behavior is AirApp-first — give the user the clickable AirApp URL. Start pnpm --dir content/kelly-writer-app dev only when local preview/debugging is explicitly requested.

Required app views (hash routes):

#/overview: metrics (needs review / approved / done / blocked), the channel breakdown, and the top drafts still needing review.
#/drafts: the review queue over channel drafts in workflow states needs_review, to_approve, changes_requested, approved, done, blocked. Each item shows a stable ref (#3), channel/format/status badges, an editable title and body, hashtags/CTA/media brief/title-options support panels, a Review note textarea, and Approve / Request changes / Block buttons that write the verdict directly onto the draft record.
#/settings: sanitized config summary — brand/audience/tone, configured channels, onboarding state, and the exact node scripts/generate_batch.mjs / node scripts/export_decisions.mjs commands to run next. Never expose secret values.

Demo mode:

?demo=1 (or ?demo=overview) opens a deterministic mock content batch ("A practical launch guide for a local-first AI workflow") for documentation and screenshots.
?demo=drafts and ?demo=settings select named mock scenes.
lang=en or lang=zh forces UI chrome language for screenshots.
Demo mode never reads or writes Busabase.

UI language: support English and Chinese chrome with Auto default. Keep titles, hooks, bodies, and hashtags in their original language.

Review Workflow

A human verdict (approve / request_changes / block / revise) writes the new status plus decision-note / decided-at (and, for approve or revise, the edited title / body) directly onto the draft record through busabase-sdk. From a standalone local preview the write merges immediately (trusted operator); from the deployed AirApp it creates a pending ChangeRequest for the trusted process to merge.

Scripts

Both scripts are trusted, skill-root Node processes with their own package.json (busabase-sdk dependency) — the AirApp browser never runs them and never calls bases.createChangeRequest/writes a local file itself.

node scripts/generate_batch.mjs --source <path-or-text> [--channels official_blog,xiaohongshu,wechat,newsletter,linkedin,x] [--audience "..."] [--cta "..."] [--source-draft-path <path>] [--apply] Reads a source (a file path or inline text), derives deterministic per-channel draft heuristics (first-pass only — Codex should improve each draft with judgment afterward, either by editing the record's title/ body or via a decision on the same record), and writes one new drafts record per channel to Busabase. Without --apply this is a dry run that only prints the drafts it would create.
node scripts/export_decisions.mjs [--apply] [--out <dir>] Re-reads Busabase for drafts with status: "approved", packages each as a Markdown file (title/channel/format/review note/body/CTA/hashtags/media brief) plus a ZIP archive (Markdown + any locally-referenced images next to the original source, resolved via source-draft-path and KELLY_WRITER_CONTENT_ROOT) under exports/<batch-id>/ at the skill root, then marks each exported draft done. Without --apply this is a dry run that only prints what would be exported.

Normal Workflow

Detect mode. Default to App UI.
Clarify or infer the source, target audience, desired channels, language, and CTA.
Run node scripts/generate_batch.mjs --source ... --apply to write a fresh batch of channel drafts to Busabase, then improve each draft's title/body with judgment (the generator's heuristics are a first pass) before handing the batch to the user.
Give the user the AirApp URL (or local preview URL) to review, edit, and approve drafts.
For a draft moved to changes_requested, re-draft it per the review comment and write it back to needs_review.
On "export approved drafts": run node scripts/export_decisions.mjs --apply to package every approved draft into a channel-ready ZIP pack under exports/ and mark it done. This skill never publishes anywhere itself — handing the exported pack to a publishing connector is a separate, explicitly authorized step.
Never export a draft without an explicit approve decision, and never re-export a draft already done.

Content Generation Rules

Preserve the source's claims. Do not invent results, dates, customer stories, statistics, prices, legal/compliance statements, or endorsements.
Ask or leave needs_review/blocked when the source lacks needed proof, product details, screenshots, links, or policy facts.
Separate platform adaptation from translation: changing channel format is allowed; changing the promise is not.
Prefer concrete hooks, specifics, and reader benefit over generic motivational copy.
Keep CTA and links consistent with the settings profile or the user's explicit request.
For Chinese-language work, support natural Simplified Chinese by default unless the source/user asks for another language.
For Xiaohongshu, produce a scroll-stopping title, short structured body, optional image/carousel brief, and hashtag set.
For long-form derivatives such as newsletter or WeChat, preserve nuance and structure; avoid shrinking the idea into slogans.
For short social posts, make each post independently understandable; do not rely on the reader seeing the original blog.

Read references/channel-playbook.md when choosing or adapting channel-specific formats.

Safety Defaults

Treat exporting as approval-required; a draft without an explicit approve decision is never eligible for scripts/export_decisions.mjs.
Store only the minimum content needed for review; keep secrets and Busabase credentials out of drafts and logs.
Keep stable ids (draft_id, ref, batch_id) so repeated updates and exports are idempotent.

Chat-Only Mode

When the user asks to avoid the UI:

Produce a compact channel plan.
Present numbered drafts with channel, title/hook, body, CTA, and notes.
Ask for approval or edits.
After approval, run node scripts/export_decisions.mjs --apply (or write the final approved pack to local Markdown directly) if the user wants files.

Never claim content is published unless the user explicitly used a publishing connector and it succeeded.

kelly-writer

Kelly Writer

App UI Screenshots

Overview