Why Your AI Coding Agent Gets Telnyx APIs Wrong (and How to Fix It With Skills)

You open your AI coding assistant and ask it to build a Python Flask app that answers an incoming Telnyx voice call and greets the caller with text‑to‑speech. The agent scaffolds the project, installs dependencies, and runs the server.

Then it crashes.

AttributeError: module 'telnyx' has no attribute 'Call'

If you’ve built APIs long enough, you already know what happened. The model guessed. It followed patterns that look right but don’t exist in the current SDK.

Without skills

# What your agent writes without skills
call = telnyx.Call()
call.answer()
call.speak(payload="Hello")

Nothing here looks unreasonable. If you were designing a voice API from scratch, those method names would make sense.

But large language models learn from training data that can be months or years old. Meanwhile APIs keep changing. SDKs get refactored. Endpoints move. Method signatures shift.

So the agent fills the gap the only way it can. It guesses.

If you're building developer platforms, this becomes a real problem. The first experience developers have with your APIis through an agent. If the agent guesses wrong, your API looks broken.

What Are Skills?

Skills solve this by giving the agent the API context it actually needs before it writes code.

A skill is a structured file that follows the Agent Skills specification. It contains verified SDK examples, endpoint structures, webhook schemas, and usage patterns.

When an agent loads a Telnyx skill, it stops relying on its training data for Telnyx code. Instead it works from the skill.

In practice that means the agent has access to:

• Correct SDK initialization patterns for the latest version
• Working code examples for API operations
• Webhook events and payload schemas
• Error handling and retry logic
• Authentication patterns

Once installed, the skill becomes part of the agent’s context. When the agent generates Telnyx code, it references the skill instead of guessing the API.

Coverage across the Telnyx platform

The repository currently includes 224 skills across 36 Telnyx products and six language environments: Python, JavaScript, Go, Java, Ruby, and curl. Beyond the core SDK skills, there are three additional plugins.

Beyond that, three additional plugins cover WebRTC clients, the Telnyx CLI, and Twilio migration.

WebRTC client plugin

For teams building calling apps in browsers or mobile apps, the WebRTC plugin includes setup instructions, call control examples, push notification handling, and quality metrics.

Twilio migration plugin

Many developers come to Telnyx from Twilio. The migration plugin scans a codebase for Twilio API calls, flags incompatible patterns, and runs smoke tests after the migration.

Telnyx CLI plugin

The CLI plugin gives agents the ability to manage numbers, configurations, and account resources through the Telnyx CLI directly from the terminal.

How to Install

Installation takes less than a minute.

Claude Code

Add the Telnyx skills marketplace:

/plugin marketplace add team-telnyx/telnyx-skills

Install the language plugin:

/plugin install telnyx-python@telnyx-skills

Replace telnyx-python with telnyx-javascript, telnyx-go, telnyx-java, telnyx-ruby, telnyx-curl, or telnyx-cli depending on your stack.

Now run the same prompt again.

With skills installed

import os
from telnyx import Telnyx

client = Telnyx(api_key=os.environ.get("TELNYX_API_KEY"))

# Answer the incoming call. call_control_id comes from the call.initiated webhook
# Learn more: https://telnyx.com/resources/what-is-call-control
response = client.calls.actions.answer(
    call_control_id=webhook_payload["data"]["payload"]["call_control_id"],
)

# Greet the caller with text-to-speech
client.calls.actions.speak(
    call_control_id=webhook_payload["data"]["payload"]["call_control_id"],
    payload="Hello, welcome to our service.",
    voice="female",
    language="en-US",
)

Same prompt. This time the code lines up with the SDK.

Windsurf

Create a .windsurfrules file in your project root and paste the relevant skill content into it. Use the same URL pattern above to fetch the skill file for your product and language.

Any Other Agent

If the agent supports the Agent Skills specification, point it to the Telnyx skills directory. If not, copy the contents of a SKILL.md file into the agent configuration.

Why This Matters

We are already seeing a shift in how developers interact with APIs. For many teams the first interface to an API is no longer documentation. It is an AI coding agent.

That changes how developer platforms need to think about documentation. Agents do not read long docs. They load context and generate code from it.

If your API does not provide structured context, the agent will try to reconstruct your API from its training data.

Sometimes it guesses right. Often it doesn’t. Skills give the agent a reliable source of truth. Instead of inventing method names, it writes code against the real API.

For developers that means less debugging. For API companies it means the difference between an API that "works" and one that looks broken on the first attempt.

Get Started

GitHub repository: github.com/team-telnyx/telnyx-skills

Works with Claude Code, Cursor, Windsurf, and any agent that supports the Agent Skills specification.

Teach your coding agent Telnyx once and it stops guessing.