Stop repeating your instructions. 🛑 Copilot now supports Agent Skills, the open standard by @AnthropicAI. Package your specialized expertise into skills once and use them everywhere. Try it in: ✅ agent mode in @code Insiders ✅ Copilot coding agent ✅ Copilot CLI Here's how to get started. 👇 github.blog/changelog/2025…

@github @AnthropicAI @code Copilot utilizes Agent Skills; instruction repetition diminishes.

@github @AnthropicAI @code Agent skills becoming a cross-platform standard is wild. Write once, run in Copilot, Claude Code, CLI. The interoperability gap between AI coding tools keeps shrinking.

@github @AnthropicAI @code Reusable skills solve the maintenance problem I hit with every agent—six months later nobody remembers why it works. Packaging logic once means my team can actually modify it when requirements shift.

@github @AnthropicAI @code We have a project where I’m required to use GitHub Copilot with different models. Now I can directly use the models’ capabilities without having to launch benchmarks for each one. This is really nice.

@github @AnthropicAI @code So we could now move our specialized instructions for vitest, playwright, ngrx/signals etc from the instructions to skills? (and adjust them a little)?

@github @AnthropicAI @code As models improve the ROI on instructions is diminishing. I feel instructions might get in the way for a powerful agent.

@github @AnthropicAI @code Maybe copilot is not geared towards coding like claude is. I have not instructed claude sonnet or opus even once. It just knows what I mean.

@github @AnthropicAI @code Agent Skills as an open standard is a huge win, the missing piece now is a local skills registry so the same skill file runs in Copilot, Claude, Vercel AI SDK, and any TypeScript agent runtime with zero rewrite.

@github @AnthropicAI @code Packaging expertise into reusable skills is how agents become actually useful.

@github @AnthropicAI @code incorporating skills seamlessly like this opens up unique innovative potentials.

@github @AnthropicAI @code Crazy

@github @AnthropicAI @code One of the most awaited features I was looking for. Now I guess I can automate the security vulnerability scanning, fixing and generating PR.. So our product is compliance with EU CRA #sbom

@github @AnthropicAI @code Long overdue. One of the things that impressed me when I moved to Cursor was having this built in. A necessary move to improve the product, not just the tooling.

@github @AnthropicAI @code agent skills are just organized prompts. mcp is just organized tool calling. the real unlock is figuring out what context to inject when. that's the harness engineering nobody talks about.

@github @AnthropicAI @code 🚀🚀🚀

@github @AnthropicAI @code Gradient descent back down the hole which you emerged

@github @AnthropicAI @code Today I transplant my copilot from instruction.md to skill.md . Looking forward to test it out

@github @AnthropicAI @code FYI @Merkousha @AmirNapster @KuomarsRahimi @mehrandvd

@github @AnthropicAI @code Bhai thoda sasta karo plan 39 dollar mahanga h yaar india me

@github @DeepLearn007 @AnthropicAI @code Agent Skills sound intriguing, especially with the potential for coherent dialogues. It’ll be interesting to see how they enhance user interactions and streamline workflows. Excited for the possibilities ahead!

@github @AnthropicAI @code VS Code is evolving so slowly compared to other open-source tools. A bit of shame is felt.

@github @AnthropicAI @code github.com/djcyphers/netw… I made a copilot-cli wrapper that leverages skills. Thank you github team!