Lanbow Claw Skill
Why do we packaged our best work as a free Skill — and what it signals about where software distribution is going.
— Lanbow
At Lanbow, we've spent years in the trenches of digital advertising — managing hundreds of campaigns, optimizing millions in ad spend, and cataloguing every mistake along the way. That accumulated expertise lives in our systems, our SOPs, our instincts.
The $10M in our headline isn't hyperbole — it's the cumulative cost of the lessons encoded in this system. Every failed A/B test, every budget misallocation, every targeting insight that only surfaces after you've burned real money. That's what we packaged.
This week, we open-sourced all of it as Lanbow Claw Skill. Free for anyone to use.
Here's why — and what it means for anyone running ads without a full team behind them.
The Ground Is Shifting
SaaS isn't going away — but it's no longer the only distribution game in town.
Vercel launched skills.sh in January and hit 8 million installs within weeks. Stripe, Supabase, Prisma, Coinbase, and Microsoft all shipped official Skills before Q1 ended. In December, Anthropic donated the Agent Skills spec to the Linux Foundation as an open standard — and OpenAI, Atlassian, Figma, and Cursor all adopted it.
NPM took a decade to reach critical mass. The Agent Skills ecosystem got there in months.
A new distribution channel has opened up alongside traditional software. The question isn't whether to explore it — it's whether you can afford not to.
Why a Skill — On Top of SaaS
We already have a SaaS product. It works. Enterprise clients use it. But we looked at the landscape and asked: is a dashboard the only way to deliver our expertise?
The answer is no.
A dashboard is great when you want to visualize data, compare campaigns, and make decisions at your own pace. But increasingly, the work happens inside an AI agent — founders and growth teams planning, executing, and iterating in conversational environments.
So we built Lanbow Claw Skill — not to replace our SaaS, but to meet users where they already are.
Distribution efficiency. A SaaS product asks users to come to you. A skill goes where they already are. When you're planning your next campaign inside Claude, ChatGPT, or Gemini — Lanbow is one command away.
Zero friction for execution. Dashboards are great for visualization. They're terrible for execution. When you need the right targeting parameters, the right budget allocation, the right creative direction — you want it delivered in the context of your work, not in a separate tab.
Trust through transparency. The Skill is open-sourced under a real license. You can read every line, audit every decision tree. That's a level of trust no SaaS demo can match.
Many Skills shipping today are wrappers — existing products with an AI interface bolted on. We went further: we took the core expertise powering our SaaS and repackaged it as a standalone, open-source Skill. It runs independently. It doesn't require a Lanbow account. It's a product in its own right.
What This Means If You're Running Ads Without a Team
Here's who we had in mind when building this.
You're a solo founder, a two-person growth team, or a small agency running paid campaigns alongside everything else. You don't have a dedicated media buyer. You don't have a strategist on retainer. You're making targeting decisions between product calls and customer support.
This isn't a niche scenario anymore. Solo-founded companies now account for 36.3% of all new businesses. An Indie Hackers survey found that solopreneurs using AI agents report 340% average revenue growth — without increasing hours worked. Sequoia has started modeling "agentic leverage" into their investment thesis.
The one-person company isn't a meme. It's a structural shift. And the question has moved from whether one person can run a full operation to which agents make it possible.
Lanbow Claw Skill is designed to be your advertising department — the same decision logic used by teams managing eight-figure budgets, delivered as a Skill your AI agent can run while you focus on building your product.
What's Inside
Lanbow Claw Skill is a single decision cycle with four connected stages:
1. Insight Report — Before you spend a dollar, understand the landscape. Feed it a URL, a business goal, audience context — get back a structured analysis: problems, opportunities, audience segments, recommended actions. Powered by real-time search, not stale databases.
2. Creative Generation — Turn insights into creative direction. Key messages, visual themes, first-draft copy, reusable prompt templates. Not final assets — a strategic starting point that would normally take a senior creative strategist hours to produce.
3. Ad Deployment — Execute campaigns directly through Meta Ads. Campaign creation, ad set configuration with granular targeting, creative binding, budget allocation — all from your agent. No manual clicking through Ads Manager.
4. Post-Campaign Re-Insight — After the spend, diagnose what happened. What worked, what didn't, and what to change for the next round. Every campaign feeds the next one.
The flow is: Observe → Decide → Execute → Iterate. The value isn't in any one stage — it's in the cycle.
What makes this different from a generic "AI ads tool" is where opinions live. Hundreds of campaigns taught us things that aren't in any API documentation: when to use Campaign Budget Optimization vs. Ad Set budgets, how to structure targeting for cold vs. warm audiences, which metrics actually predict performance vs. which ones just feel good. Those judgments are encoded in the system — not as suggestions, but as validation rules that prevent costly mistakes before they happen.
How Expertise Gets Encoded (Not Just API Calls)
You don't need to read code to use the Skill. But if you're wondering what separates $10M of encoded expertise from a weekend wrapper around Meta's API — here's what's actually under the hood.
Lanbow Claw Skill is 5,291 lines of TypeScript with only 3 runtime dependencies: Commander.js, cli-table3, and Zod. Everything else is hand-written. The architecture was designed specifically for AI agent execution: zero global state, isolated dependency graphs per command, no side effects between calls. Any LLM can invoke any command without worrying about shared state.
The real value lives in what we call business rules as code — validation logic that encodes expensive lessons so you never have to learn them yourself:
Budget mutual exclusion. Meta's API lets you set both daily and lifetime budgets on a campaign. It shouldn't — it silently ignores one. We catch this locally, instantly, before your money is at risk.
Bid strategy requirements. Pick a COST_CAP bid strategy without specifying a bid amount? Meta's API returns a cryptic error 30 seconds later. We validate it in milliseconds with a clear message.
Creative mode conflicts. Three different ways to define ad creatives in Meta's system. Mix them — even slightly — and you get undefined behavior. Our schema enforces exactly one mode, every time.
These aren't edge cases. They're the kind of rules that cost real money when you discover them through trial and error — which is exactly how we discovered them.
One more detail worth mentioning: Meta's video upload API operates on an entirely separate domain, uses a different content type, and requires authentication in a non-standard way. None of this is obvious from the docs. Our video system handles files up to 4GB with automatic chunked uploads and exponential backoff retries. And Meta's error code 200? It means "permission denied", not "OK." We learned this through production incidents. You don't have to.
For those who want to go deeper: the full source is on GitHub. Read every line, audit every decision.
The Business Decision Behind "Free"
Let's address the obvious question: why give away your best work?
Because we've done math on what matters.
Traditional SaaS acquisition works like this: run ads → drive to landing page → get a demo request → qualify the lead → close the deal. Every step has friction. Every step loses people. Customer acquisition costs keep climbing.
Skill-based acquisition works like this: someone installs a Skill → runs their first insight report → sees the quality of the output → trusts the expertise → scales up → needs more.
The funnel is shorter. Trust is earned through usage, not a pitch deck. And the cost of distribution is essentially zero — we're not competing for browser tabs and login slots. We're competing for a line in someone's Skill config.
This aligns with where the industry is heading. Meta itself is planning to enable fully AI-automated advertising by end of 2026 — where advertisers submit a product image and budget, and AI handles the rest. The advertising stack is being disaggregated. Expertise that sits behind a dashboard is at risk. Expertise that runs natively inside AI workflows compounds.
We'd rather have 10,000 teams running our Skill than 100 leads in a CRM.
What's Coming: The Layer Above Execution
Lanbow Claw Skill handles the execution cycle for individual campaigns — observe, decide, execute, iterate. But execution is one layer.
The hardest problem is what happens across campaigns: how capital gets allocated between markets, how creative fatigue gets detected and managed as a system variable, how the lessons from this quarter's campaigns inform next quarter's strategy automatically. That's the layer where individual campaign optimization becomes compounding growth infrastructure.
We've been building this layer for our enterprise clients — teams managing $20M+ in annual ad spend — and we'll be writing about it in depth: how reinforcement learning meets institutional memory, how creative direction becomes governed output instead of reactive guesswork, and why the companies that build decision infrastructure outperform the ones that optimize individual campaigns.
The skill is where you start. The system behind it is where it gets interesting.
The Bigger Picture
Here's what we think every company — and every solo operator — with domain expertise should be asking:
Can your core capability be packaged as a skill?
Not "can you add an AI chatbot to your product." Not "can you expose an API." But: can you take the hard-won knowledge that makes you valuable and encode it into something an AI agent can execute?
If the answer is yes, the distribution math has changed. Ignoring Skills as a distribution channel in 2026 is like ignoring the App Store in 2009. You're not competing for attention through ads and SEO anymore. You're competing for a slot in someone's AI workflow. The barrier to trial just dropped to zero.
We're going all in. And we're starting by giving away our best work.
FAQ
Is this actually free? What's the catch?
No catch. Lanbow Claw Skill is fully functional, open-source, and doesn't require a Lanbow account. Use it, inspect it, build it. As your needs grow — deeper analytics, multi-market governance, dedicated support — that's where Lanbow's enterprise offering comes in.
Do I need to be technical to use it?
No. The Skill runs inside your AI agent (Claude, ChatGPT, etc.). You describe what you want in plain language — "analyze my competitor's ad strategy" or "create a campaign targeting US millennials interested in fitness" — and the agent executes the right commands.
What platforms does it support?
Currently Meta Ads (Facebook + Instagram). The four-stage cycle — Insight, Creative, Deployment, Re-Insight — is designed to extend to additional platforms.
Lanbow Claw Skill is open-source and available now. Install it, run your first insight report, and see what $10M worth of advertising expertise looks like — working inside your AI agent, not behind another login screen.
