golem-annotate-agent-ts

To add metadata to TypeScript agents for LLM tool discovery, you apply class-level @description and method-level @prompt/@description decorators. This annotates agent purpose and method execution guidance for intent matching.

The correct way to use decorators for TypeScript agent method prompting is applying @prompt and @description directly to methods. This provides execution guidance and behavioral documentation, while avoiding invalid agent options like embedding description in @agent().

An LLM fails to select the right method on a TypeScript agent class when missing method-level @prompt annotations. Without discovery metadata, the system cannot match user intent to specific method execution guidance.

No, you cannot put description text inside the @agent() decorator for TypeScript agents. Embedding description in @agent() is an invalid agent option; class-level descriptions must use the separate @description decorator.

You need to annotate TypeScript agent classes with prompt and description metadata when maintaining agents with multiple methods, like an InventoryAgent, that an LLM must correctly understand and invoke based on user intent.