@khemaridh: Skills are becoming a problem. Not because they don't work. But because they're so effective. I'm now seeing teams (on …
Summary
The author highlights problems with effective AI skills in teams, such as poor sharing, version control, and management, leading to outdated and inefficient practices.
View Cached Full Text
Cached at: 08/23/26, 07:45 AM
Skills are becoming a problem. Not because they don’t work. But because they’re so effective.
I’m now seeing teams (on the buyside) where individuals have a 5-20 skills. They work and they’re used often.
But here’s the problem.
Skills are still stuck in single-player mode.
How do you share a skill? Email me the .md file. (And piss off, if it’s a zip file.)
Then the models change and Claude recommends shrinking skills by 80%. The power users do it. No one else received the memo. (Duh, they’re not on X all day.)
Some skills are centralized. Do we really need 20 review-nda skills? One person should own this, make sure it’s always dialed-in. But who is that person? How do you ensure that everyone has the most up-to-date skill? (And what happens if that person leaves?)
Other necessary items: Version control, ownership, search-ability and discovery, descriptions. Pretty easy stuff, but not available in the current harness.
And don’t get me started on testing. Right now it’s all based on vibes. (I’m guilty of this too.) It’s hard to run evals on qualitative knowledge work.
Don’t tell me git. That proves you’ve not understood how the average knowledge worker operates. They will NEVER use git. EVER.
Who’s building the solution?
@mstockton shared this for OpenAI
Super cool! Can this interact directly w/Cowork? That’s where all my clients live
Oh this is cool. The comments are making me realize that I haven’t fully grok’d the plugin as a delivery vehicle
Cool, checking it out!
Similar Articles
@op7418: https://x.com/op7418/status/2065232309310427565
This article discusses the concept of Skills in the AI agent ecosystem, arguing that Skills are more than prompts—they are packaged capabilities that externalize human expertise into reusable workflow units. The author shares design principles and case studies from building popular Skills.
@svpino: Skills are the most important development in AI over the last 6 months. I think it was Anthropic the one that invented …
Skills are highlighted as the most important AI development in recent months, with Bolt being the latest tool to support them, allowing teammates to share and stack skills automatically.
I used to be proud of these skills. Now AI agents do them better.
The author reflects on how AI agents now outperform them in code navigation, debugging, and report drafting, and asks others about experiences with multi-agent workflows like MCP, Anvita Flow, and Agent Protocol.
We keep adding “skills” to our agents and have no idea which ones actually work. Solved problem?
A PM at an internal developer platform highlights the challenge of tracking which AI agent skills are actually invoked and effective, and asks the community if there are existing tools or solutions for this observability problem.
Stop building AI workflows your team is afraid to touch
The article highlights common issues with complex AI workflows, such as lack of documentation and knowledge transfer, and suggests practices like versioning and testing to improve team handoffs.