Six weeks ago, Salesforce and Anthropic announced Claudeforce, a partnership that puts Salesforce's own CRM data inside Claude as a plugin, shipping with 37 prebuilt sales skills for things like meeting prep, deal health review and pipeline updates. That's not a startup experimenting with a chatbot bolt-on. That's the world's biggest CRM vendor deciding Claude Skills are worth building an entire product line around.
At the same time, Gartner published research saying that by 2028, AI agents will outnumber human sellers ten to one, and fewer than 40% of sellers will say those agents actually improved their productivity.
Both things are true at once, and that's really the whole story. Skills are having a real moment in sales. Most of what's showing up because of that moment isn't going to help anyone.
What's actually happened this year
Anthropic's Skills feature lets Claude load a folder of instructions, scripts and reference material only when a task calls for it, rather than stuffing everything into one giant prompt. It works across the Claude apps, Claude Code and the API, on Pro, Max, Team and Enterprise plans.
Claudeforce is the biggest sign yet that this idea has left the "interesting technical feature" stage. Announced on 26 August 2026, it goes in both directions. Claude gets embedded across Salesforce's own products, and Salesforce shows up inside Claude as a plugin carrying real customer data and business logic, with what Anthropic and Salesforce call "governed actions" so a rep can update a pipeline or review a deal without leaving the conversation. It's in pilot with select customers now, with an open beta expected in September 2026 and more skills planned for later in the year.
Marc Benioff called it "the world's #1 AI meets the #1 AI CRM." Dario Amodei framed it as pointing Claude at "the customer information and business context that they've been building in Salesforce for decades." Whatever you think of the phrasing, the underlying fact is hard to argue with. When a company the size of Salesforce restructures a chunk of its product roadmap around Claude Skills, that's not hype cycle noise. That's a vendor betting real engineering time on the idea sticking.
The marketplace side of the story is messier
Away from Salesforce's own announcement, the broader Claude Skills ecosystem has grown fast in a way that's harder to feel good about. One tracking report puts it plainly. The skills marketplace landscape went from a single registry in December 2025 to eight major marketplaces by the second quarter of 2026, ranging from a few hundred curated listings up to marketplaces claiming hundreds of thousands of skills scraped in from GitHub.
Marketplaces, six months apart
Source. Agentman, "The Agent Skills Ecosystem in 2026."
More listings isn't the same as more usable skills. The same report found a security audit turned up over 140,000 issues across the skills it tested, averaging more than six problems per skill, and that roughly a third of tested skills carried some form of prompt injection vulnerability.
Catalog size, in other words, is close to a meaningless number on its own. A marketplace with 800,000 skills and a marketplace with 20 hand-reviewed ones are answering completely different questions, and only one of those numbers tells you anything about whether a given skill actually does what it claims.
Gartner's number is the one worth sitting with
The ten-to-one prediction gets the headlines, but the detail underneath it is more useful than the ratio itself. Gartner's research, based on a survey of 210 chief sales officers and senior sales leaders run in January and February 2026, found that CSOs who make a genuine overhaul of their data, automation and rep experience are five times more likely to see real ROI from AI than those who bolt on a quick fix.
Gartner VP Analyst Dan Gottlieb summed it up in one line, saying more agents "will not automatically mean more productivity." That's the same pattern showing up twice in the same research from two completely different angles. More skills on a marketplace doesn't mean more useful skills. More agents in a sales org doesn't mean more productive sellers. Both problems come from the same place. It's now trivially easy to ship something that looks like an AI sales tool, and much harder to ship one that was actually built by someone who understands the job it's supposed to do.
Where that leaves someone actually choosing a skill
None of this means the trend is fake. Salesforce didn't build Claudeforce for a fad, and a market moving from one registry to eight in six months is a real signal, not a mirage. It means the interesting question has already shifted from "should I try a Claude Skill for sales" to "how do I tell a real one from the other 140,000 issues sitting in a marketplace somewhere."
Worth checking before you install anything.
- Was it built by someone who's actually done the job it's coaching or automating, or does it read like a generic prompt template with a sales-shaped label stuck on it?
- Has it been tested against a real rubric more than once, and does the seller say so plainly, including where it's failed?
- Is a subscription actually buying you something ongoing, like hosted infrastructure or live data, or is it a fixed piece of logic being rented back to you every month for no real reason?
We built rAInmaker with exactly that bar in mind, and we'd apply the same questions to it if we were the ones buying. It's worth trying the free Lite version against your own outreach before you take our word for any of it.
Sources. Salesforce Newsroom, "Salesforce and Anthropic Announce Claudeforce" (26 Aug 2026) · Apex Hours, "Claudeforce Explained" · Anthropic, "Introducing Agent Skills" · Agentman, "The Agent Skills Ecosystem in 2026" · Gartner Newsroom, "Gartner Predicts AI Agents Will Outnumber Sellers 10 to 1 by 2028" (28 Jul 2026).