skill-wechat-publisher · WeChat Article Creation & Publishing
WeChat official-account article creation and publishing: given a topic, reference article, or document, it automatically researches the web, writes, illustrates, formats, and de-AI-checks the article — then publishes to the account's draft box (never auto-sends).
The core is anti-AI-detection writing: skeleton draft → humanizing rewrite pass (9-item mandatory checklist) → AI-flavor self-check gate (blocks publishing at score ≥45). Supports multiple account voices (distinct styles per account are themselves an anti-detection strategy) and 15 formatting themes. Also offers a picture-message mode (5–10 card wall + short captions).
Example invocation: "Turn these notes into a WeChat article and send it to drafts."
Full brief
Positioning
skill-wechat-publisher is a full "material → WeChat draft box" pipeline: research, writing, illustration, formatting, anti-AI checks, publishing. It delivers to the draft box — never auto-sends.
Core capabilities
- Web research: authoritative layer (news/data/cases/official reports) + human layer (community quotes as raw material); key figures cross-checked across sources.
- Structured writing: 8 article structures and 7 opening hooks to choose from; skeleton drafts may carry AI flavor.
- Humanizing rewrite pass: a dedicated pass against a 9-item mandatory checklist (sentence variety, AI-phrase blacklist, fact density, punctuation variety, etc.).
- Anti-AI detection gate: ai_score runs automatically before publishing; ≥45 blocks the release; flagged passages are rewritten at the sentence level until they pass.
- Multi-account voices: built-in dual account voices (AI-product / engineering), rewritten with clearly distinct styles; Easel profile tone overrides when available.
- Formatting: Markdown → WeChat HTML with 15 themes + inline highlight markup.
- Dual publishing channels: default backend session (QR sign-in, no app_id/app_secret, no IP allowlist); optional fallback to the official HTTP API (credentials + IP allowlist required).
- Picture-message mode: WeChat "image message" style — 5–10 card wall + 100–300 word captions.
Workflow
- Understand the need, gather material (collect human-flavor raw material: real names/dates/amounts/versions/pitfalls)
- Web research (
research.md, every item sourced) - Skeleton draft → humanizing rewrite pass
- Generate illustrations (needs image API key; external images or skip otherwise)
- Convert format and apply theme
- AI-flavor self-check gate (no pass, no publish)
- Publish to draft box (title/abstract user-confirmed; backend session, never auto-send)
Inputs & outputs
| Input | Notes |
|---|---|
| Topic / material | Required — reference articles, text, or documents |
| Target account | Optional — sets voice and theme; backend-session mode publishes to the signed-in account |
Output: article draft in the WeChat draft box (title, abstract, cover, formatted body).
Fit
- Turning topics/documents/notes into WeChat articles fast
- Accounts sensitive to AI detection needing human-feeling writing
- Multi-account content production and formatting
Before you start
- QR-sign in to the WeChat backend first (backend-session mode needs no app_id/app_secret or IP allowlist); official API mode needs credentials + IP allowlist.
- Image generation needs an image API key (skippable or external when unavailable).
- Dry-run by default; formal publishing only after the user confirms title, abstract, and account.
skill-wechat-publisher is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.