Skills are the most flexible extension mechanism — any Python function can become an Agent tool.

Built-in Skills

Skill	Function	Description
Current Time	get_current_time	Get current date and time
Calculator	calculator	Precise math calculations
Text Analysis	analyze_text	Count words, lines, paragraphs
Unit Converter	unit_convert	Length, weight, temperature conversion

Adding a New Skill

Step 1: Write a Python function

Edit modules/skills/tools.py:

def my_new_skill(param1: str, param2: int = 10) -> str:
    """Tool description — the Agent uses this to decide when to call it.
    param1: description of param1
    param2: description of param2, defaults to 10
    """
    result = do_something(param1, param2)
    return f"Result: {result}"

Key requirements:

• The docstring is critical — the Agent decides when to invoke the tool based entirely on it
• Parameters need type annotations (str, int, float, etc.)
• Return type is str
• Describe when to use and parameter meanings in the docstring

Step 2: Register in get_all_skills()

def get_all_skills() -> list:
    return [
        FunctionTool.from_defaults(fn=get_current_time),
        FunctionTool.from_defaults(fn=calculator),
        FunctionTool.from_defaults(fn=analyze_text),
        FunctionTool.from_defaults(fn=unit_convert),
        FunctionTool.from_defaults(fn=my_new_skill),   # add here
    ]

No other files need to change. Restart and the new skill is available.

Knowledge Skills

Beyond tool functions, DataMind also supports knowledge skills — Markdown documents placed in data/skills/ that are indexed and searchable:

data/skills/
├── database-ops-sop.md
└── code-review-guide.md

The Agent uses skill_search to find relevant procedures and best practices from these documents.

How the Agent Decides

1. Receives user question
2. Reads all tool descriptions (docstrings)
3. LLM judges which tool best matches the intent
4. Extracts parameters and calls the tool
5. Integrates tool output into the final answer