Frequently asked questions

Skills Wiki is an AI skill management platform — a single place to find, enable, and manage the capabilities your AI assistant has. Instead of manually copy-pasting data into Claude or ChatGPT, you connect a skill once and your AI can pull live data from tools like Shopify, Semrush, Google Analytics, and HubSpot directly.

MCP (Model Context Protocol) is an open standard, published by Anthropic, that lets AI assistants connect to external tools and live data. Skills Wiki is a managed MCP platform — it runs the infrastructure so you don't have to host your own MCP server or write integration code.

Claude (Desktop and Claude.ai), ChatGPT (Custom GPTs and web), Gemini, and any other assistant that supports an MCP connection. You manage your skill stack in one dashboard and it stays in sync across every platform you connect.

Skills Wiki's Evolution Loop monitors every skill continuously. When an external API changes and a skill breaks, an AI supervisor detects the failure and patches it — usually within 24 hours, without any action from you.

Yes. The Basic plan is free and includes unlimited skill activations and the community marketplace. Pro unlocks the full curated skill catalog and external service connections. See the pricing page for current details.

A short intake form (about 3 minutes) followed by a hands-on review of your setup — up to 45 minutes of our time. You get back a 1-page written diagnosis and a short video walkthrough naming what we found and what's at risk if it isn't addressed. It's free, comes with no obligation, and nothing is ever auto-applied to your store.

We review what you told us and take a look at your setup or site. You'll get a 1-page diagnosis doc plus a short Loom video, typically within two business days. If there's a real finding worth acting on, the doc ends with an offer to book a 25-minute call — no pitch you didn't ask for. If there's nothing to fix, you still get the full doc and a warm close.

No — most businesses that reach out have already done that themselves. We start with an audit of what's actually happening inside your AI tools today, then build a safely-scoped Skills Wiki combo for the specific task that's broken, manual, or too risky to automate as-is.

Three shapes: Audit only (written findings, no implementation — for when you want to understand the problem first), Audit + fix (we implement a scoped combo addressing the findings), and Ongoing oversight (a recurring health-check that catches the next silent failure before it compounds). Pricing is per-project since setups vary too much for a flat rate.

It's founder-operated, not outsourced to an agency. The same person who reviews your setup builds and maintains Skills Wiki itself — and runs every process on our own stack first, before it goes anywhere near yours.

The common failure mode isn't the AI model being wrong — it's a process gap. No dry-run to preview the change before it's live, no rollback if something breaks, and no scoping of exactly what "bulk" touches. A tool that changes 10,000 product listings at once with no safety net will eventually change the wrong 500.

Usually not — but it's worth being precise about what actually failed. Often the incident itself gets resolved, but the response is a blanket ban on automation, even the safe cases. That trades a one-time cost for a permanent one: doing a task manually forever. The fix is narrower than "never automate" — it's adding the oversight layer (dry-run, rollback, scoping) that was missing the first time.

Three things, minimum: can you preview the change before it's live (dry-run), can you undo it if something's wrong (rollback), and is the blast radius scoped to exactly what you intend to touch, not your whole catalog by default. If any of those is missing, that's the gap to close before running it again.

Off-the-shelf automation apps apply the same process to every customer. A combo built through the diagnosis is configured specifically for what your business does and doesn't want touched — including what stays manual and what gets a rollback step, based on what the audit actually found.

No. The diagnosis is a review only — we look at your setup and describe what we found. Nothing is auto-applied. Implementation only happens as a separate, explicitly scoped and agreed engagement.