Skip to main content
The Agent Skills system provides several built-in tools that allow your AI agents to dynamically interact with, load, and execute the content of Skills during a session. These tools empower agents to read the instructions defined in a skill, or even execute scripts that are bundled alongside a skill’s SKILL.md file.

​
Available Skill Tools

There are two primary tools that handle skill operations:

​
1. activate_skill

Loads the full instructions and content of a specialized skill. An agent will use this when it determines that the user’s request matches a skill’s description, allowing it to “page in” the necessary context. Input Parameters:
  • skill_id (string, required): The UUID of the skill to activate.
Output:
  • Returns the full content (SKILL.md) of the skill wrapped in XML-like tags, informing the agent of the virtual directory where the skill’s files reside.

​
2. read_execute_skill_file

Allows the agent to read an attached file OR execute a script from a skill’s bundle. If the agent asks to read a text file or markdown file from the skill’s virtual directory, this tool reads and returns the text. If the agent asks to execute a Python script (.py), this tool safely executes the script within the UBIK logic node environment and returns the output. Input Parameters:
  • skill_id (string, required): The UUID of the skill.
  • file_path (string, required): The relative path of the file (e.g., scripts/run.py or docs/info.md).
  • args (array of strings, optional): Command-line arguments to pass if executing a python script.
Output:
  • status: Success or error.
  • action: Either 'read' or 'executed'.
  • content: The text content (if read).
  • stdout / stderr: The standard output and error streams (if executed).
  • artifacts / generated_files: Arrays containing any files, images, or data generated by the script during execution. These files can be retrieved using the GET /assets/tools/{tool_id}/{execution_id}/{filename} endpoint.

​
How Agents Use Skill Tools

When you attach a Skill to an Agent, the agent is automatically made aware of the skill’s name and description.
  1. Trigger: The user asks a question (e.g., “Analyze this dataset using our standard methodology”).
  2. Activation: The agent realizes it has a “Data Analysis Methodology” skill. It calls activate_skill with that skill’s ID to retrieve the detailed steps.
  3. Execution: The skill’s instructions might tell the agent to run a specific script. The agent then calls read_execute_skill_file with the path scripts/analyze.py to perform the work.
  4. Completion: The agent reads the stdout of the script and formulates a final response for the user.