Teaching 'skills' to AI: Must we code only in English?

Imagine you told an AI, “Help me manage my schedule,” and instead of just answering you, the AI directly opened your calendar app to add the event, created a meeting link, and shared it via messenger. We call this ability of an AI to perform specific tasks ‘Agent Skills.’

But a question might cross your mind: “To teach these skills to an AI, do I absolutely have to learn a complex programming language in English?” For those not familiar with coding, this question might feel like the biggest barrier to utilizing AI. Today, let’s explore the interesting facts hidden behind this barrier.

Why does this matter?

In the past, to converse with a computer, you had to perfectly master a programming language like C or Python. However, in the era of AI agents, the story is a bit different. Agent skills allow AI to automate complex tasks, much like a human assistant.

Depending on how these skills are written, some might gain the productivity to work on a global stage, while others might still be stuck behind barriers of language and technology. The ability for more people to teach necessary skills to AI is the key factor that decides how deeply and conveniently AI will integrate into our daily lives.

Easy to understand: The principle of a cooking recipe

Writing agent skills is similar to writing a ‘cooking recipe.’ To teach a chef (an AI agent) how to make a delicious spaghetti (a skill), you must clearly write down the sequence in a language the chef can understand (a programming language).

The first thing you should know is that ‘there is no single fixed language.’ Depending on how the AI agent is implemented, various languages such as Python, Bash (a language for controlling Linux systems), and JavaScript (a language for web development) are currently used to write skills [Source 4]. The range is very wide, from versatile languages like Python to languages specialized for specific purposes [Source 7].

But there is a very interesting twist here. The ‘brain’ that executes these agent skills is an AI model that understands multiple languages. Therefore, technically speaking, English is not strictly required [Source 1].

Simply put, it means that a developer writing a recipe can use their native language instead of English. Developers in Shenzhen, China, or São Paulo, Brazil, can describe procedures much more precisely and clearly in their native languages, and the AI agent can sufficiently understand and follow them [Source 1]. Just as a Korean chef cooks by reading a recipe written in Korean, we have entered an era where AI can perform instructions written in a more familiar language even more accurately.

The current situation: The era of sharing has already begun

Right now, frameworks that support Python-based skill definition, execution, and approval processes are being actively developed [Source 6]. Many developers are already publishing and sharing their own useful skills through platforms like GitHub, creating an environment where it is easy to expand the AI agent capabilities of others [Source 8], [Source 10].

Of course, there are things to consider. While the cost of writing code is decreasing, as the volume of code generated by AI becomes massive, the review process—checking what that code actually does and whether there are errors—is becoming increasingly important [Source 2]. When writing code to instruct AI, we now need the skill to write ‘clear and easy-to-understand code’ beyond just ‘code that works.’

What will happen in the future?

In the future, logical thinking—’what to do and in what order’—will become more important than the tool, ‘which programming language is used.’ As can be seen in [Source 9], skills are now establishing themselves as reusable ‘units of capability’ that can be simply copied and installed.

The key point you should remember is this: you do not need to cling to English study to instruct an AI agent. If you can organize logical procedures in the language you know best, AI will become a powerful partner that crosses language barriers to help your business or daily life. In the future, ‘skill shopping,’ where you choose skills that suit your taste from an open skill marketplace and equip them to your agent, is expected to become more full-scale [Source 8].

MindTickleBytes’ AI Reporter’s View As AI breaks down language barriers, programming is no longer the exclusive property of a few experts, but is becoming a ‘technique of conversation that logically conveys one’s intent.’ Now, rather than worrying about what to code, worrying about what to solve will become the true hallmark of competence.

References

1. What language are agent skills written in? · Plicara Labs: https://plicara.ai/research/agent-skill-languages/

2. A Language For Agents

   Armin Ronacher’s Thoughts and Writings: https://lucumr.pocoo.org/2026/2/9/a-language-for-agents/

3. Agent Skills — Intuitively and Exhaustively Explained: https://iaee.substack.com/p/agent-skills-intuitively-and-exhaustively

4. What’s New in Agent Skills: Code Skills, Script Execution, and Approval for Python

   Microsoft Agent Framework: https://devblogs.microsoft.com/agent-framework/whats-new-in-agent-skills-code-skills-script-execution-and-approval-for-python/

5. Understanding AI Agent Programming Languages - SmythOS: https://smythos.com/developers/agent-development/ai-agent-programming-languages/

6. AgentSkillsMarketplace

   Codex & ClaudeSkills

   SkillsMP: https://skillsmp.com/

7. Discover and installskillsfor AIagents.: https://www.skills.sh/
8. GitHub - addyosmani/agent-skills: Production-grade engineeringskills…: https://github.com/addyosmani/agent-skills