Skip to content
Get packs

Guide · Agent skill

How to use the PicoClaw Agent Helper agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The picoclaw-agent skill is a set of instructions that teaches an AI agent how PicoClaw actually works: its CLI commands, config.json keys, file paths, provider setup and workflows. This guide is for people who may never have used an agent skill before. By the end you will know what the skill does, how to install it, how to ask for help in a way that triggers it, and what to check before you trust the output.

PicoClaw Agent Helpersipeed/picoclaw · MIT

Quick start

  1. 1Install PicoClaw. Make sure the picoclaw command works on your machine before using the skill.
  2. 2Download and unzip the skill. Download the ZIP from this page and unzip the picoclaw-agent folder.
  3. 3Place the skill folder. Copy it to ~/.claude/skills/ for personal use, or to a project's .claude/skills/ folder. Alternatively upload it under Settings → Capabilities → Skills in the Claude apps.
  4. 4Ask a PicoClaw question. Describe your task and name PicoClaw, for example a model, MCP server, cron job or gateway problem.
  5. 5Review what the agent writes. Check config.json and any MCP env files for keys and settings before saving.
  6. 6Verify with PicoClaw. Run picoclaw status or picoclaw mcp list --status and compare the output with what the agent described.

What the picoclaw-agent skill does

An agent skill is a folder containing a SKILL.md file of written instructions. When a skill is active, the agent reads those instructions and follows them for matching tasks. This one covers PicoClaw itself: its CLI, config, gateway, authentication, models, skills, MCP servers, cron jobs, session routing, self-evolution, built-in slash commands and repository internals.

The point is accuracy. With the skill active, the agent uses PicoClaw's real commands, config.json keys, file paths and provider setup steps when you ask about PicoClaw, rather than guessing. The skill is a reference and workflow guide. It is made of text instructions and command lists and includes no bundled scripts.

Install the skill

Download the ZIP from the skill's page on this directory and unzip it. Then choose one of these options.

For a personal setup, put the unzipped picoclaw-agent folder in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

  • Personal: ~/.claude/skills/picoclaw-agent
  • Project: .claude/skills/picoclaw-agent
  • Claude apps: Settings → Capabilities → Skills

Trigger the skill with natural requests

You do not need a special command. The skill's description tells the agent to use it for tasks about PicoClaw CLI commands, config.json, gateway, auth, models, skills, MCP servers, cron, routing, sessions, self-evolution, slash commands or repository internals. Ask in those terms and name PicoClaw, so the agent knows which tool you mean.

Good requests name the goal and the context, for example: help me add a local Ollama model to my PicoClaw config.json, or why does my MCP server not show up in picoclaw mcp list.

  • Ask about a specific task: add a cron job that posts a daily summary.
  • Name the file or command you are working with: check my config.json model_list.
  • Say what went wrong: picoclaw gateway starts but my chat channel is silent.

Examples of what the agent can help with

The skill carries the commands the agent needs. The first-run flow is picoclaw onboard, which initialises ~/.picoclaw/config.json and ~/.picoclaw/workspace. Model work uses picoclaw model and picoclaw model add. Skills use picoclaw skills list and picoclaw skills install. MCP servers are managed with picoclaw mcp add, picoclaw mcp list and picoclaw mcp test.

Cron jobs are a common request. The skill notes that the current CLI supports recurring jobs only, using --every or --cron, so the agent will not promise a one-shot job flag.

picoclaw onboard picoclaw model add -b http://localhost:8000/v1 -k dummy -m my-model -n local picoclaw cron add --name "Daily summary" --message "Summarize today's logs" --cron "0 18 * * *"

What you need before you start

The skill assumes PicoClaw is already installed, so the picoclaw command must work on your machine. You also need credentials for the model provider you choose, such as an OpenAI or Anthropic account, or a local endpoint such as Ollama, which needs no API key.

Node.js and npx are needed only if you follow the example MCP server commands, which use npx.

  • PicoClaw CLI installed (the picoclaw command)
  • A provider API key or account, or a local endpoint
  • Node.js/npx, only for the example MCP server commands

Safety notes: scripts and network

The skill runs no scripts and makes no network calls of its own. It is text the agent reads.

The main risk is secrets. The guide tells the agent to write API keys into config files and MCP env files. Before you save anything the agent writes, read it and check where the key ends up. Do not paste real keys into a chat if you would not want them in that conversation.

  • Runs scripts: no
  • Uses network: no
  • Review config.json and MCP env files before saving

Tips for better results

Ask the agent to check its work against your real setup. Commands can behave differently for other tools, so the skill is only useful for PicoClaw. If you ask for something PicoClaw does not do, such as a one-shot cron flag, the agent should say so rather than invent one.

Use picoclaw status and picoclaw mcp list --status to give the agent real output to reason from. Note that picoclaw mcp only edits configuration; servers are loaded later by the host or gateway.

  • Paste real output from picoclaw status or picoclaw mcp list --status.
  • Remember picoclaw mcp does not keep servers running on its own.
  • Skill loading priority is ~/.picoclaw/workspace/skills, then ~/.picoclaw/skills, then builtin skills.

What people use it for

  • Onboarding a new PicoClaw installation with picoclaw onboard and a first provider login.
  • Adding a local Ollama or OpenAI-compatible endpoint to model_list.
  • Adding an MCP server with picoclaw mcp add and checking it with picoclaw mcp test.
  • Scheduling a recurring job with picoclaw cron add using --every or --cron.
  • Debugging a gateway that starts but does not deliver messages to a chat channel.

Frequently asked questions

Is the picoclaw-agent skill free?

The skill is published under the MIT licence, as listed on this page. PicoClaw itself is also the project of sipeed/picoclaw. Any model provider you connect may charge for its API, which is separate from the skill.

What do I need to use the picoclaw-agent skill?

You need the PicoClaw CLI installed so the picoclaw command works, plus credentials for your model provider or a local endpoint. Node.js and npx are needed only if you use the example MCP server commands.

How do I install the picoclaw-agent skill?

Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.

Does the skill run scripts or access the network?

No. The skill contains only a SKILL.md text file with instructions and command lists, and it runs no scripts and uses no network. It does tell the agent to write API keys into config files, so review those files before you save them.

Can I use the skill without installing PicoClaw?

The skill is designed for people who have PicoClaw installed. Its commands assume PicoClaw is present and may give wrong results for other tools, so install PicoClaw first.

How is this different from a general AI assistant?

The skill only applies to PicoClaw tasks. It makes the agent use PicoClaw's real commands, config.json keys and paths, instead of general guesses, and it does not help with unrelated topics.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you are setting up, troubleshooting or extending a PicoClaw installation, or contributing to its codebase, and you want an agent to use real PicoClaw commands and config keys. Skip it if you want a general-purpose assistant or do not run PicoClaw, since it only applies to PicoClaw tasks. Always review any API keys the agent writes into config files before saving.

Download the skill