Custom Skills for Hermes Agent: Full Tutorial

Key Concepts

• Hermes Agent: An AI-powered agent framework that utilizes "skills" to perform specific tasks.
• Skills Hub: A centralized repository for discovering and installing pre-configured community or official skills.
• Skill Structure: The architectural requirement for custom skills, consisting primarily of a skill.md (metadata/instructions) and a scripts/ folder (executable logic).
• Qwen 3.7 Max: The LLM model used in the demonstration for its high performance and cost-efficiency.
• Kubernetes: An orchestration framework used as a case study for pre-built skill integration.
• Autonomous Skill Creation: The capability of the Hermes Agent to generate its own skill files based on natural language instructions.

---

1. Installation and Setup

To begin using Hermes Agent, the user must perform a terminal-based installation.

• Command: A single-line installation command is used to initialize the environment.
• Configuration: During setup, the user selects a custom direct API (e.g., Alibaba Cloud Model Studio) and specifies the model (Qwen 3.7 Max).
• Gateway: The agent can be installed as a background service to ensure persistent availability.
• Verification: Once installed, the hermes command launches the interactive chat interface.

2. Creating Custom Skills

Creating a custom skill involves a specific directory structure within the .hermes/skills/ folder.

• Directory Structure:
  • [Category Name]/[Skill Name]/: The root folder for the skill.
  • skill.md: Contains the skill description, purpose, and usage instructions for the agent.
  • scripts/tools.py: Contains the actual Python code/logic to be executed.
• Example (Stock Price Skill):
  • The skill.md instructs the agent to use Yahoo Finance.
  • The tools.py script defines a function that accepts a stock symbol and returns the price in USD.
• Validation: After creating the files, the user runs hermes skills list to confirm the skill is detected locally. The agent then successfully retrieves real-time data (e.g., Apple stock price) by invoking the custom script.

3. Installing Pre-built Skills

Users can extend the agent's capabilities by importing community-developed skills.

• Process: Search the Skills Hub, copy the provided CLI command, and execute it in the terminal.
• Security: The system distinguishes between "trusted" and "untrusted" skills. Community skills may require explicit permission or force-installation if they are not pre-verified.
• Case Study (Kubernetes): By installing the Kubernetes skill, the agent can query cluster health, providing detailed metrics on node resource usage, CPU, and memory consumption.

4. Autonomous Skill Creation

Hermes Agent can generate its own skills through natural language prompting.

• Methodology: The user provides a prompt (e.g., "Understand about Praveen AI help CLI command and then create a skill based on that").
• Advantage: While the agent automates the creation of skill.md and script files, the user should understand the manual structure to validate the agent's output for accuracy and security.
• Advanced Components: Beyond basic files, skills can include:
  • Additional Docs: Supplementary documentation.
  • Templates: Output formatting rules.
  • Assets: Necessary supplementary files.

5. Management Commands

The hermes skills CLI provides a suite of tools for lifecycle management:

• list: View all installed skills.
• create/update/delete: Manage the local skill library.

---

Synthesis and Conclusion

The Hermes Agent framework relies on a modular "skill" system that allows for both manual development and autonomous generation. By separating the descriptive metadata (skill.md) from the functional logic (tools.py), the system remains highly extensible. The ability to integrate external APIs and CLI tools—demonstrated through the Stock Price and Kubernetes examples—transforms the agent from a simple chatbot into a functional automation tool. The recommended workflow is to master the manual creation process first to ensure the ability to audit and validate the skills generated autonomously by the agent.