A simple no-fuss local LLM agent implementation with skills support

1. Supports any OpenAI compatible API. (local or remote)
2. File and directory listing tools for reading codebases and documents
3. Automatic skill discovery (project and user-scope)
4. Automatic skill activation by LLM
5. Manual skill activation (using /)
6. Python script calling support in skills

1. Clone the repo
2. Run uv sync or install the dependencies using pip
3. Run main.py. (or uv run main.py for uv)

Adding new tools is easy. Just add a new entry in tools.py.

1. Create the tool function (any python function will do).
2. Create a Tool object and pass it the schema of the tool.
3. Add the Tool object to tools list.
4. Restart the agent.

Adding new skills is even easier!

1. Create a new folder for the skill in ./.agent/skills (project-scope) or ~/.agent/skills (user-scope).
2. Add the SKILL.md file. (it has to be named exactly like this).
3. Add yaml frontmatter to the skill (name, description are mandatory, others optional)
4. Add any other folders, files or scripts the agent may need for this skill.
5. Restart the agent.