Skip to main content
cat
Credits: 🤖 Agent Z
Agent Skills help agents pull just the right context from simple Markdown files. The agent first sees only a skill’s name and short description. When a task matches a skill, the agent loads the rest of SKILL.md, follows the steps, and can execute commands in a shell environment as specified by the skill.

Agentor x Skills

Agentor provides native support for skills. Simply provide the directory or path to a SKILL.md file, and Agentor will automatically inject the skill metadata into the system prompt. Key features of Agentor skills:
  • Starts light: Discover skills by name and description only
  • Loads on demand: Pull full instructions from SKILL.md when relevant
  • Executes safely: Run skill-driven commands in an isolated shell

Creating Your First Skill

Skill follow the following layout:
Follow these steps to build and integrate a custom skill into your agent.
1

Create the skill directory

Organize your skill in a dedicated folder. At minimum, you need a SKILL.md file.
2

Define instructions in SKILL.md

The first few lines should contain a clear name and description. The rest of the file should provide detailed, step-by-step instructions for the agent.
SKILL.md
3

Register the skill

Pass the path to your skill folder when initializing your agent.

Usage Examples

Creating a GIF

Best Practices

  • Clear Descriptions: The first 2-3 lines are critical. They determine if the agent “activates” the skill.
  • Modular Steps: Break complex logic into smaller, numbered instructions.
  • Error Handling: Tell the agent what to do if a command fails (e.g., “If ffmpeg is missing, suggest the user install it”).
Use the scripts/ folder for complex Python or Bash scripts that the agent can call via the shell. This keeps the SKILL.md focused on logic rather than raw code.
Security Note: Since skills can execute shell commands, always run your agents in isolated environments (like Docker or a VM) when using skills from untrusted sources.
Last modified on December 19, 2025