NocoBase Blog Writing (AI & No-Code Section)
This skill is derived from a sentence-by-sentence statistical analysis of all 147 articles (2024-2026, with both English and Chinese versions) in the official blog's "AI & No-Code" category. The goal is to produce SEO-friendly articles that readers can't tell were written by AI and that are ready for publication.
Use articles from 2026 as your benchmark; articles from 2024 are a textbook example of what to avoid. The writing style within the same column changed significantly over two years:
| Metric (English) | 2024 | 2026 |
|---|---|---|
| Hollow adjectives (seamless, robust, comprehensive, powerful, streamline, leverage…) per 1,000 words | 6.35 | 0.33 |
| Median article length | 1,986 words | 2,706 words |
| Articles with a comparison table | 38% | 63% |
| Articles with an FAQ | 0 articles | 21/43 articles |
The same is true for the Chinese version: in 2024, words like "powerful," "flexible," "seamless," "easy," and "efficient" were ubiquitous; by 2026, they had almost completely disappeared.
Four Non-Negotiable Rules
- Don't make up anything. The persuasive power of this column comes from real information: real Reddit/HN/X discussions (with links), GitHub Stars checked on the day of writing, publicly available prices from the official website, and features documented in the product docs. Go look things up (GitHub API, official website, docs, WebSearch) before you write. If you can't find a source, mark it with
【TO_VERIFY: …】and leave it for a human. AI writing fails most often here; one fabricated fact destroys the entire article's credibility. - Specific first, abstract second. Every judgment must be immediately followed by a concrete example: a real business object (customer, contract, work order, invoice), a real role (salesperson, finance, project manager), a real workflow (
Lead → Opportunity → Quotation → Closed deal). Replace vague adjectives with "what can it do, for whom, and in what scenario." - Be measured and fair when writing about competitors. First, objectively explain what the other product is good at, then discuss differences and limitations. When recommending NocoBase, attach conditions (e.g., "If your goal is X, you could start by looking at NocoBase"). Avoid self-proclaimed terms like "best" or "No.1."
- Distinguish clearly between NocoBase 3 and NocoBase 2. "NocoBase" in an article refers to NocoBase 3 by default—the "Agent-oriented development platform" as termed on the official website: Agents write code on the Core, and Studio is responsible for governance. NocoBase 2 is the "AI + No-Code development platform": interface configuration with AI employees. The capabilities, terminology, links, and pricing of the two versions are completely different. When writing about topics like no-code, visual building, or AI employees, explicitly write "NocoBase 2." See section 2 of
references/nocobase-facts.mdfor the rules.
Workflow
1. Clarify the Task
Confirm the following items before starting to write. When the user only provides a topic, infer reasonable defaults, list them in 3-5 lines for the user to confirm, then proceed. Do not ask a series of follow-up questions.
- Primary keyword and search intent: What are readers searching for? What problem are they trying to solve?
- Article type: A Open-source tool roundup / B Alternatives & Comparisons / C Build-it-yourself tutorial / D Concept explainer (see table below)
- Target audience: Developers, IT managers, business teams, outsourcers/integrators
- Language: English, Chinese, or both (the Chinese and English versions should have the same structure but with localized headings and phrasing—see
references/voice.md) - NocoBase angle and version: In this article, what specific reader problem does NocoBase solve? Which capabilities are used? Is the article about NocoBase 3 (default) or NocoBase 2 (no-code, AI employee topics)? For product terminology, features, and links, refer to
references/nocobase-facts.md - Opening material: One real community discussion or industry phenomenon. If none exists, go find one (Reddit, Hacker News, X, Google Trends)
2. Research
- Deduplicate first: Search
references/existing-posts.mdfor the primary keyword and its synonyms (e.g., retool, appsmith, crm, ticket). If an old article on the same topic exists, follow the rules at the top of that file (typically suggesting a rewrite at the original URL) and include the plan in the delivery notes. Without deduplication, a new article will cannibalize the ranking of the old one. - Use WebSearch to search for the primary keyword. Look at what the top-ranking articles cover and what they miss. We need to fill in the gaps they leave unclear (real scenarios, costs, permissions, and long-term maintenance).
- Find opening material: Priority is given to real community discussions. For backup sources when Reddit is inaccessible, see "General Openings" in
references/article-types.md. - Fact-check each product you will mention: official website, docs, pricing page, GitHub repo, Stars (use
https://api.github.com/repos/<owner>/<repo>to getstargazers_count; write it as22.7kand note the query date in the delivery notes). - Select 8–15 relevant older articles from
references/internal-links.mdfor internal links.
3. Outline
Create an outline based on the skeleton for the corresponding type in references/article-types.md. Use the fixed H2/H3 headings from that document. If the article will exceed 2,000 words, show the outline to the user first before writing.
4. Writing
Write according to references/voice.md. Before writing, read the "Rewrite Comparison" section there; it's faster than memorizing the rules to get a feel for the style.
Example sentences from the reference files are taken from published articles. Learn from their structure and tone, but do not copy them directly into your new article. Repeating this practice will make the entire blog sound like it's using the same template.
5. Self-Check
- Run
python3 scripts/check_article.py <article.md>(the script path is relative to this skill's directory) and fix the issues it reports: filler words, overly long paragraphs, heading and description lengths, missing sections. - Go through the checklist in
references/seo-checklist.md. - Read only the first three paragraphs and the last two paragraphs. Ask yourself: Does this sound like someone who has actually done this work? Could any sentence still be valid if you replaced the product name with another? If so, rephrase it so it is only applicable to this specific article.
6. Delivery
Output a single Markdown file with the following format:
---
title: <SEO title, 50-60 characters>
slug: <lowercase-hyphenated-keyword>
description: <meta description, 120-155 characters>
primary_keyword: <primary keyword>
secondary_keywords: [<2-5 keywords>]
category: AI & No-Code
author: <author, format consistent with the official blog, e.g., Deng Lijia, Yuan Tianqi, Zhang Lu>
---
# <H1, can be different from title but must contain the primary keyword>
<Body text>
Appendix after the body: a "Delivery Notes" section (not for publication), including:
- A list of screenshots needed: file names are English descriptive words (e.g.,
nocobase-crm-opportunity-board.png) + alt text. - All
【TO_VERIFY】items, along with the query dates for GitHub Stars and pricing. - If both language versions are required, suggested titles and descriptions for the other language.
The Four Article Types
| Type | Title Pattern (Real Examples) | When to Use |
|---|---|---|
| A. Open-Source Tool Roundup | Top 8 Open Source AI Assistant Tools by GitHub Stars / After Claude Code: 6 Open-Source Tools You Should Know | For a hot topic, popular AI tools, or top-starred projects under a specific GitHub topic. |
| B. Alternatives & Comparisons | How to Choose a Smartsheet Alternative: 7 Tools Compared / NocoBase vs Baserow: Flexible Databases vs Enterprise Systems | When readers are already using or evaluating a product and need alternatives or want to make a decision. |
| C. Build-it-Yourself Tutorial | How to Build a Production-Ready CRM with AI and NocoBase / How to Design an IT Asset Management System: Data Model, Lifecycle, and Workflows | To demonstrate "AI + NocoBase" from requirements to deployment with a specific business system. |
| D. Concept Explainer | What Is AI No-Code? A Practical Guide to No-Code Platforms in the AI Era | For new concepts and trends, where readers want to understand the basics before choosing a tool. |
For the complete skeleton, fixed subheadings (in both English and Chinese), and paragraph structure for each type, see references/article-types.md.
Tone Essentials (For details, see references/voice.md)
- Short paragraphs: The median paragraph length in English is about 20 words; 1-3 sentences per paragraph. Use single-sentence paragraphs for transitions (e.g.,
But code is not the same as an application.). Average sentence length is about 15 words. - Pronouns: Use "we" for editorial actions ("we selected," "we came across"), and "you" when speaking to the reader. First-person stories in the opening can use "I."
- Measured language: Use phrases like "better suited to," "relatively," "should still confirm whether…" Avoid: seamless, robust, powerful, comprehensive, game-changer, unlock, empower, delve, In today's fast-paced world, In conclusion.
- Chinese: Avoid terms like "赋能" (empower), "助力" (assist), "无缝" (seamless), "强大" (powerful), "轻松" (easy), "打造" (build), "一站式" (one-stop), "降本增效" (reduce costs and increase efficiency), "综上所述" (in conclusion). Use "比如" or "例如" for examples.
- Emojis for functional marking only: 💬 column description, 💡 further reading, 🎉 product announcement, 🚀 try online, ⭐/❤️ share at the end of the article. Do not use for decoration in the body text.
- Use dashes sparingly: No more than 1 em dash per 1,000 words in English. Use commas, colons, or periods instead.
Where NocoBase Fits in the Article
- The column description line (the 💬 line) is placed after the opening scenario but before the first H2 of the body text. See
references/nocobase-facts.mdfor the exact text. - In roundups and comparisons, NocoBase is listed first and gets a deeper write-up, but the heading structure is consistent with other products. Do not use special formatting.
- When writing about NocoBase, anchor it in specific capabilities and scenarios. No slogans. For NocoBase 3, common points: pre-built components let the Agent write only the business logic; a unified project structure and development standards; changes are code (reviewable, rollbackable); business Agents; Studio; self-hosting. For NocoBase 2, common points: visual building, workflow canvases, AI employees. See section 6 of
references/nocobase-facts.mdfor the full list. - Follow the official terminology: in English, "agent" is lowercase; "business agent" (NocoBase 2 calls it "AI employee"). In Chinese, write "Agent" or "业务 Agent", not "智能体" or "代理".
- In the FAQ, include 1-2 questions that recommend NocoBase based on usage scenarios. The very last question can be directly about NocoBase (e.g., "Is NocoBase only for enterprises?").
Reference Files
| File | When to Read |
|---|---|
references/article-types.md | Before outlining, read the section for the corresponding type. |
references/voice.md | Before writing; before rewriting someone else's draft. |
references/seo-checklist.md | When finalizing the title, slug, and description; during the pre-delivery self-check. |
references/nocobase-facts.md | When writing about NocoBase, placing CTAs and product links. |
references/existing-posts.md | First step of research: check for old articles on the same keyword. |
references/internal-links.md | When selecting internal links and "Related reading" at the end of the article. |
scripts/check_article.py | Run before every delivery. |