Hello Hermes

Problem

Hello Hermes is a minimal smoke-test skill for validating an Agent Skills publication loop: whether a skill can be created automatically, exposed through SKILL.md, and later used to steer a model toward structured output. It is not a business skill; the point is to exercise the basic cycle of publication instructions, stepwise field collection, and JSON emission. It is especially useful for end-to-end checks across a publishing platform, a model gateway, and a skill loader.

How It Works

According to the materials, it is the first Skill published automatically through Hermes Agent. Its runtime behavior is intentionally narrow:
- The skill instructions ask the model to request fields step by step.
- The model collects user-provided values in the requested order.
- The final response merges those values into a JSON object rather than free-form prose.

This makes it useful as a template or regression example: first confirm that the skill loads correctly, then check whether the model follows the structured-output constraint.

Boundaries

It covers only a minimal example. Do not treat it as a full Agent capability. If the task requires real business rules, multi-step tool calls, permission checks, or external system access, the SKILL.md needs to be extended with the corresponding tools, validation, and execution policies.