---
title: "Discussion on \"From Tool Calling to Skill-Based LLM Execution\" | Hashnode"
description: "Discussion on \"From Tool Calling to Skill-Based LLM Execution\". PS: Traditional tool-calling agents waste up to 90% of their context window on JSON schemas. There's a better way.\nThe Solution\nInstead of defining 50 tool schemas, describe capabilities in plain English and let the LLM generate executable Python cod..."
url: "https://hashnode.com/posts/from-tool-calling-to-skill-based-llm-execution/6969da9e354e6012352906e9"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [From Tool Calling to Skill-Based LLM Execution](https://karthikeyakollu.hashnode.dev/from-tool-calling-to-skill-based-llm-execution?utm_source=hashnode&utm_medium=feed)

PS: Traditional tool-calling agents waste up to 90% of their context window on JSON schemas. There's a better way. The Solution Instead of defining 50 tool schemas, describe capabilities in plain English and let the LLM generate executable Python cod...

Jan 16 · 3 min read · karthikeyakollu.hashnode.dev  
Read the article

-   [#llm](https://hashnode.com/tag/llm)
-   [#context-engineering](https://hashnode.com/tag/context-engineering)
-   [#ai](https://hashnode.com/tag/ai)
-   [#ai-engineering](https://hashnode.com/tag/ai-engineering)

## Written by

[KOLLU KARTHIKEYA](https://hashnode.com/@KarthikeyaKollu)

@KarthikeyaKollu · Visakhapatnam, India

  
[View profile](https://hashnode.com/@KarthikeyaKollu)

Full Stack Web Developer

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
