How to use the MiMoCode Product Docs agent skill (mimocode-docs)

What an agent skill is and what this one makes the agent do

An agent skill is a folder of instructions that an AI agent loads when a request matches it. The skill's SKILL.md file describes when to use it, and supporting reference files hold the detailed material. The agent reads only the files it needs for the question.

The mimocode-docs skill covers MiMoCode features, TUI and CLI commands, keybindings, agent modes (build and plan), configuration, file locations, providers, models, authentication, and custom OpenAI-compatible or Anthropic-compatible API endpoints. It can also help inspect existing config, make minimal changes, and verify them. Its scope is MiMoCode only.

• Answers how-to questions such as switching agent modes or setting up a model provider from the reference files.
• Explains where config and data live on disk.
• Helps make small, checked edits to mimocode.json or mimocode.jsonc files.
• Explains how other tools or subprocesses can borrow this instance's models through the /v1 endpoints and mimo llm-server tokens.

Install the mimocode-docs skill

Download the skill ZIP from this page and unzip it. For a personal install, place the mimocode-docs folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/mimocode-docs/
• Project: .claude/skills/mimocode-docs/ inside your project

Requirements

The skill's answers are about the MiMoCode CLI, which provides the mimo binary. The commands and config it describes assume that binary is installed. The skill does not run scripts and does not use the network.

• MiMoCode CLI (mimo binary) installed on the machine where you work.
• No scripts run by the skill.
• No network use by the skill.

Trigger the skill with natural requests

You do not need a special command. Ask the agent a normal question about MiMoCode and the skill loads when the request matches its description. Requests that name a base URL, API key, model ID, provider, or mimocode.json file are especially likely to trigger it.

Questions about entering or leaving plan mode also trigger it. The skill notes that the user enters a mode, and the agent does not switch into plan mode on its own.

• How do I switch MiMoCode from build to plan mode?
• Set up a custom OpenAI-compatible base URL for MiMoCode in my project config.
• Where does the global mimocode.jsonc file live?
• My MiMoCode TUI lags over SSH. What should I check?
• How do I let a subprocess use this instance's models through /v1?

Examples of useful questions

Good questions name the goal and the place you are working. For config changes, ask the agent to read the current file first and make only the change you need.

The reference files cover commands, config layout, providers, permissions, MCP sampling, the capability API and dynamic workflows, so a question can point to any of these areas.

• Ask for the purpose of a command, for example what mimo models lists.
• Ask for a minimal change, for example adding one permission rule to .mimocode/mimocode.json.
• Ask for a check, for example whether a provider entry is valid before you use it.

Safety notes before you let the agent edit config

The skill can change mimocode.json files. Those files may hold real API keys, and the agent could expose or mistype credentials. Review every config change before you accept it, and keep keys out of prompts where you can.

The capability API gives a subprocess a scoped token rather than your provider key. Treat any listening port as something to check before you expose it.

• Review diffs to mimocode.json or mimocode.jsonc before accepting them.
• Do not paste real API keys into chat if you can avoid it.
• Do not use the skill for tools other than MiMoCode; its answers may be wrong outside that scope.