OpenClaw skills and plugins
OpenClaw skills and plugins are how the Gateway gains new abilities, from posting to a channel to calling an API to running a recurring job. This page explains what they are, where they live on the host, how the marketplace fits in, and how to read and manage them from your iPhone.
TL;DR
OpenClaw skills and plugins extend what the Gateway can do. A plugin adds a capability to the Gateway, such as a messaging channel or a tool integration. A skill is a unit of behavior the agent can call to complete a task. Both live as files in the OpenClaw config directory on the host, which means you can install, edit, and audit them like any other code. Onepilot reaches that directory over SSH, so you can browse the skill files with syntax highlighting, diff changes in the git tab, and schedule skill runs with cron, all from the phone. The agent runs on the host, and you keep full sight of what it can do.
Read and manage OpenClaw skills on the host, straight from your iPhone over SSH. Download it on the App Store.
What are OpenClaw skills?
OpenClaw skills are units of behavior the agent can invoke to get something done. A skill bundles instructions, and often code, that teach the agent how to perform a specific task in a repeatable way. When the agent decides a task matches a skill, it loads that skill and follows it rather than improvising from scratch.
Skills live as files in the OpenClaw config directory on the host. Because they are files, you can read exactly what a skill tells the agent to do, edit it, version it in git, and remove it. That transparency matters for a shell-capable agent, since a skill is the difference between a vague instruction and a precise, auditable procedure the agent runs.
What is the difference between an OpenClaw skill and a plugin?
A plugin adds a capability to the Gateway itself, while a skill teaches the agent a task. Plugins are the integration layer: a channel plugin connects the Gateway to Telegram, Discord, or Slack, and a tool plugin gives the agent access to an external API or service. They expand what the Gateway is wired into.
Skills sit one level up. They use whatever capabilities the plugins expose and chain them into a procedure the agent can repeat. A useful way to hold it: plugins decide what the agent can reach, and skills decide what the agent knows how to do with that reach. Both are files on the host, so both are inspectable and reversible.
Where do OpenClaw skills come from?
OpenClaw skills come from three places: the ones shipped with the framework, ones you install from a marketplace or repository, and ones you or the agent write directly. The marketplace route lets you add a published skill in one step, which is the fastest way to give the Gateway a new ability without authoring it yourself.
Whatever the source, an installed skill ends up as files in the OpenClaw config directory. That is the key point for trust: a marketplace skill is not a black box, it is code on your host that you can open and read before you let the agent use it. Onepilot's file browser is built for exactly that kind of pre-flight check.
How do I manage OpenClaw skills from my iPhone?
Onepilot connects to the OpenClaw host over SSH, so the skill files are directly in reach. The file browser shows the config directory with syntax highlighting across more than 20 languages, so you can open a skill, read what it does, and confirm it before the agent relies on it. The git tab shows real diffs when a skill changes, which is how you catch what a self-editing agent altered.
From there you can do the operational work too. Use the terminal to install or remove a skill, and use cron to schedule a recurring skill run, such as a daily summary or a weekly cleanup, without writing crontab syntax by hand. The agent keeps running on the host while you inspect and adjust from the phone.
FAQ
What are OpenClaw skills?
OpenClaw skills are units of behavior the agent can invoke to complete a task. A skill bundles instructions, and often code, that teach the agent how to do something repeatable. Skills live as files in the OpenClaw config directory on the host, so you can read exactly what each one tells the agent to do, edit it, version it in git, and remove it.
What is the difference between an OpenClaw skill and a plugin?
A plugin adds a capability to the Gateway, such as a Telegram, Discord, or Slack channel or an external tool integration, while a skill teaches the agent a task that uses those capabilities. Plugins decide what the agent can reach; skills decide what the agent knows how to do with that reach. Both are files on the host, so both are inspectable and reversible.
Where are OpenClaw skills stored?
Installed OpenClaw skills end up as files in the OpenClaw config directory on the host where the Gateway runs. Because they are files, a marketplace skill is not a black box; it is code on your machine that you can open and read before letting the agent use it. Onepilot's file browser reaches that directory over SSH for exactly that kind of review.
How do I add a skill to OpenClaw?
You can add an OpenClaw skill from a marketplace or repository in one step, or write one yourself, and the skill lands as files in the OpenClaw config directory. Onepilot lets you do the operational side from the phone: use the terminal over SSH to install or remove a skill, then read the resulting files in the browser to confirm what the agent will run.
Can I manage OpenClaw skills from my iPhone?
Yes. Onepilot connects to the OpenClaw host over SSH and gives you a syntax-highlighted file browser to read skill files, a git tab to see real diffs when skills change, a terminal to install or remove them, and cron to schedule recurring skill runs. The agent keeps running on the host while you inspect and adjust from the phone.
Manage OpenClaw skills from your phone
Download Onepilot on the App Store.
See also: What is OpenClaw?, Install OpenClaw on any server, OpenClaw setup & keep-alive, Run OpenClaw on iPhone.