I Built 67 Config Validators and Made $0
Last month I went all-in on ClawHub, the agent skills marketplace. The plan: find empty niches, build config validators for every popular tool I could think of, price them at $49-99, profit.
I built 67 of them. tsconfig.json, helm values.yaml, pyproject.toml, Docker Compose, nginx, Terraform, CircleCI, you name it. Each one a standalone Python script, zero dependencies. Feed it a config, get back a list of issues.
Total revenue after a month: $0.00.
The embarrassing part
ClawHub doesn't have a payment mechanism. Everything on the registry is free. My $49-99 pricing tags were just metadata that nobody ever saw.
I could have found this out in five minutes on Day 1. I found it out on Day 67.
Empty niches aren't always opportunities
I kept getting excited about zero-competition scores. Terraform validator? No competitors! nginx validator? Nobody!
Turns out, nobody was competing because nobody was searching. Developers who use nginx already have nginx -t. TypeScript users have the compiler. ESLint users have ESLint. I spent weeks building tools that compete with built-in features of the tools they're supposed to validate.
What people actually install
The top ClawHub skills get 10K-35K downloads. They're all agent capabilities: things like Capability Evolver (35K), WhatsApp integration (16K), browser automation (11K). Not standalone scripts. Integrations that let an AI agent do something it couldn't do before.
My validators were Python scripts with a SKILL.md stapled on top. They didn't use AI reasoning. They didn't connect to anything. They just... ran a linter.
What I'm doing now
The 67 validators stay up as portfolio proof. But I've shifted to building actual agent capability skills. So far I've published 12:
Sentry Error Triage, Slack Messaging, Codebase Onboarder, Dependency Health Check, ADR Generator, Migration Safety Checker, Error Diagnosis, Git Changelog Generator, Prompt Evolution Engine, PR Review Assistant, Incident Response Runbook, and Task Retrospective.
The difference: these teach an agent how to reason about a problem domain, not just run a script. The Codebase Onboarder doesn't just count files, it produces a structured guide for someone landing in an unfamiliar repo. The Migration Safety Checker doesn't just parse SQL, it evaluates lock implications and rollback risk.
I also bundled the 20 most generally useful validators into a DevOps Config Linter Suite with a unified runner and put it on Lemon Squeezy for $19. If validators are going to live anywhere, a single "lint everything" tool makes more sense than 67 separate downloads.
What I'd tell past me
Check whether the platform takes money before building 67 things to sell on it. An empty niche on a marketplace can mean "untapped market" but it usually means "nobody wants this." And if you're building for an AI skills marketplace, the product is new agent abilities, not wrappers around existing CLI tools.
The validators weren't a total waste. I learned the publishing workflow, the skill format, and how ClawHub's search and discovery works. But three good integrations would have taught me the same things and actually gotten installs.
All skills are free on ClawHub. The linter bundle is on Lemon Squeezy.