Google ka official Agent Skills repo kya hai? AI agent ko skills kaise sikhayein
Agar aap AI tools ke saath roz kaam karte ho, jaise ChatGPT, Claude, Gemini, Cursor ya koi coding agent, to ek problem aati hai: model purani jaankari se jawab deta hai, ya Google ke products ke steps galat bata deta hai. Iska ek jawab Google ne April 2026 me diya: ek official "Agent Skills" repository.
Short answer: Agent Skills chhoti Markdown files hoti hain jo kisi technology ya kaam ka compact instruction manual hoti hain. Google ne ye skills github.com/google/skills par rakhi hain. Ye Google Cloud Next 2026 ke Day 1 (22 April 2026) par launch hua tha, aur shuru me 13 skills ke saath aaya.
Skill kya hai?
Skill ko aise samjhiye jaise AI agent ke liye ek chhota, focused documentation. Google ke blog ke mutabik skills Markdown me likhi jaati hain, aur inme reference files, code snippets aur dusre assets ho sakte hain. Agent ko zaroorat padne par hi wo information load karta hai, poora documentation ek saath nahi.
Google ne ek problem ka naam liya hai: "context bloat". Bahut zyada documentation context window me bhar di jaaye to model confuse ho sakta hai aur token cost badhti hai. Skills isi ka jawab hain.
Repo me kya hai?
Launch ke time repo me 13 skills thi, jo in par focus karti thi:
- Products: AlloyDB, BigQuery, Cloud Run, Cloud SQL, Firebase, Gemini API aur Google Kubernetes Engine (GKE)
- Well-Architected pillars: Security, Reliability aur Cost Optimization
- Recipes: Onboarding, Authentication aur Network Observability
Google ke mutabik aage aur skills is repo me aayengi.
Install kaise karein?
Google ke blog me ye command di gayi hai:
npx skills install github.com/google/skills
Ye skills Antigravity, Gemini CLI aur dusre third-party agents ke saath use ho sakti hain. Install karne se pehle apne agent ki documentation dekh lein ki wo skills ko kaise load karta hai, kyonki ye har tool me ek jaisa nahi hota.
Khud test kaise karein
- Ek chhota kaam lo, jaise Cloud Run par ek simple service deploy karna.
- Pehle bina skill ke agent se steps poochho. Phir skill install karke wahi sawal dobara poochho.
- Dono jawab compare karo: steps, commands aur error handling. Jo sahi lage wahi rakho.
Ye ek aam experiment hai, koi official benchmark nahi.
Students ke liye matlab
AI seekhte waqt sirf prompt likhna kaafi nahi hota. AI ko sahi context dena zaroori hai, aur skill files isi ka practical example hain: ek technology ka sar-sar summary nahi, balki kaam ke liye zaruri steps.
Ek baat dhyan me rakhein: ye skills Google Cloud products ke liye hain. Aapka product kisi aur platform par hai to ye seedha kaam nahi karengi.
Official sources:
— Devesh Sarkaar, Founder, Aivesh Academy
Join our WhatsApp channel: https://whatsapp.com/channel/0029Vb8x2347YSd7f4EUue3q
Related posts
Written by Devesh Sarkaar, founder of Aivesh Academy. More AI Tools posts · All posts