中文 | English
Give your AI agent real business sense. business-model-craft-skill enables any agent to analyze a company's business model using the Business Model Canvas framework — producing bilingual reports, structured data, and a locally previewable web page.
Most AI-generated business analyses share the same flaw: you finish reading and can't remember a single concrete claim.
They're full of lines like "the company is customer-centric," "faces intense market competition," and "operates a sustainable business model" — each sentence sounds plausible, but together they say nothing.
business-model-craft-skill flips that. It uses the Business Model Canvas as a forcing function — pushing the analysis to answer real questions: How does this company actually make money? Is the moat real or just a head start? Where are the risks nobody's talking about? The writing is modeled after Stratechery, Not Boring, and Harvard Business Review: opinionated, sourced, and worth finishing.
The hardest part of building this skill wasn't the code — it was the thinking methodology behind good business analysis.
Here's the thing about how people analyze companies: staring at a company's homepage and trying to "figure it out" rarely works. What works is a framework — something that turns an open-ended essay into a structured fill-in-the-blank. That's when you start finding the interesting parts.
The Business Model Canvas is exactly that kind of tool.
Introduced in 2005 by Alexander Osterwalder — co-founder of the strategy consulting firm Strategyzer — the Canvas breaks any business down into 9 building blocks. Nine questions a company has to answer about how it operates, creates value, and makes money.
These nine blocks aren't independent. They push against each other. And as a company moves through different stages, the canvas shifts too.
I've written a deep-dive guide on how to actually use the Canvas: Mastering the Business Model Canvas: A Deep Dive Guide . I also built a tool around it — businessmodelcraft.com — a free canvas builder with a library of real company case studies. This skill is the next step: bringing that same methodology to AI agents.
business-model-craft-skill distills the core ideas of the Business Model Canvas, along with two years of hands-on experience building around it, into a single skill — so your agent can reason about business models the way a sharp analyst would.
One prompt. Four outputs.
Name a company, or let the skill pick one from today's market movers. When the conversation wraps up, you'll have:
| File | Description |
|---|---|
[slug].html |
A local case page — double-click to open, built-in Chinese / English toggle |
[slug].zh.md |
Chinese deep-dive: six chapters, sharp opening take, numbers with sources, real conclusions |
[slug].en.md |
English version, written in the style of WSJ / HBR |
[slug].json |
Bilingual structured BMC data, field-compatible with businessmodelcraft.com |
Research you can trust
Every analysis is grounded in SEC filings, earnings call transcripts (Seeking Alpha), founder interviews, and reporting from WSJ and Bloomberg. Every number has a citation.
The agent also runs a mandatory AI-filler cleanup pass before saving — scrubbing the output of clichés, hollow transitions, and the kind of language that makes business writing unreadable. The goal is prose that reads like Stratechery or Not Boring, not like a generated summary.
Option 1: Let your agent install it
Works with Claude Code, OpenClaw, or any agent that can read URLs. Just send it this:
Install this skill: https://github.com/hengchilde/business-model-craft-skill
Option 2: Download manually
Download the repo, unzip it, and tell your agent the folder path to install from.
- Open the repository page
- Click the green
<> Codebutton on the right - Select Download ZIP
- Unzip the file — that's your Skill folder
Using the skill
Once installed, just start a conversation with your agent. Ask it to analyze a company's business model, or explicitly tell it to use the business-model-craft-skill. That's it.
Reports are saved locally. Priority order:
- A path you specify in the conversation
- A global output folder you've set via environment variable (
BUSINESSMODELCRAFT_OUTPUT_BASEorBUSINESSMODELCRAFT_OUTPUT_DIR) - Default:
~/Documents/BusinessModelCraft/[slug]/
When done, the agent will tell you the exact path and a file:// link you can click to open.
- Python 3.8+ (for the scripts — the agent handles dependencies automatically)
- An internet connection (for research and market data)
Optional: Alpha Vantage API Key
The stock-picking script pulls from Google Finance and StockAnalysis by default — no key needed. Alpha Vantage is an optional fallback for candidate validation. Sign up at alphavantage.co with just an email address; the free tier gives you 25 requests a day. Once you have a key, ask your agent to add it to the skill config.
MIT