Skills Guide

Superpowers Skills on OpenClaw: What Ports and What Does Not

Q: Does Superpowers officially support OpenClaw?

No. Superpowers documents installers for fourteen agents including Hermes Agent, and OpenClaw is not among them. There is no official OpenClaw plugin installer.

Q: Can I use the skills anyway?

Yes. Superpowers skills are plain SKILL.md files, and openclaw skills install accepts git: refs and local skill directories.

Q: What does not carry over?

Plugin-level features do not carry over: slash commands like /superpowers:brainstorm are Claude Code plugin features, and workflow enforcement comes from the host agent. On OpenClaw the skills load as documents rather than as enforced workflows.

Q: Where do installed skills land?

In the agent workspace at ~/.openclaw/workspace/skills/ /SKILL.md. Use --global to install into the shared managed skills directory instead.

Q: How do I check a skill is actually loadable?

Run openclaw skills check to see which skills are ready, visible or missing requirements, and openclaw skills list --eligible to show only usable ones.

Superpowers is an MIT-licensed agentic skills framework with official installers for fourteen agents. OpenClaw is not one of them — but the skills themselves are ordinary SKILL.md files, and OpenClaw can install those from git. This page is about doing that honestly, including the parts that genuinely do not transfer.

Start from the accurate version

Plenty of write-ups imply that any agent can “install Superpowers”. That is not what the project claims. It publishes a per-agent installer for fourteen runtimes, each using that runtime's own plugin mechanism, and OpenClaw has no entry. Anyone telling you to run openclaw plugins install obra/superpowers is guessing.

What is true is narrower and still useful: the framework's value lives in SKILL.md documents, and OpenClaw's skill system reads exactly that format.

Install a skill from git

OpenClaw 2026.7.1 ships a skill installer that takes three kinds of reference: a ClawHub slug, a git repository, or a local directory.

# The three accepted forms
openclaw skills install @owner/slug        # ClawHub
openclaw skills install git:<repo>         # any git repository
openclaw skills install ./path/to/skill    # local directory

Because Superpowers keeps each skill in its own folder under skills/, the practical route is to clone the repository once and install the specific skills you want as local directories, giving each an explicit slug:

git clone https://github.com/obra/superpowers.git
openclaw skills install ./superpowers/skills/writing-skills --as writing-skills

--as matters here: without it you inherit whatever slug the directory implies, and two skills from different packs can collide. --force overwrites an existing workspace skill, and --global installs into the shared managed directory rather than one agent's workspace.

Verify before you trust it

openclaw skills list --eligible
openclaw skills check
openclaw skills info <name>

check is the one that saves time. A skill can be installed and still not be usable because a requirement is missing, and that shows up as the agent quietly not using it rather than as an error you would notice.

What you get, and what you do not

	On Hermes Agent	On OpenClaw
Install path	Official one-line plugin install	Per-skill via `skills install`
Skill content	Full set	Full set, chosen skill by skill
Slash commands	Provided by the plugin	Not available
Enforced workflow hand-offs	Yes, plugin-driven	Advisory — the agent loads the document
Updates	Through the plugin	Re-install after a `git pull`

The honest summary: on OpenClaw, Superpowers becomes a well-written library of methodology documents rather than a system that makes the agent follow them. That is worth having, but it is not the same product experience, and if the enforced workflow is what you want then Hermes Agent is the supported route.

The alternative worth knowing about

OpenClaw has its own capability in this space. openclaw skills workshop manages pending skill proposals — the built-in Skill Workshop can draft a skill from a conversation that worked, scan it, and queue it for your approval. If your goal is “my agent should get better at the things I actually do” rather than “my agent should follow a published methodology”, that path fits OpenClaw better than porting a framework built for another runtime.

Either way, OpenClaw Launch gives you a managed instance with a workspace you can install skills into, without provisioning a server first.

Superpowers and OpenClaw FAQ

Does Superpowers officially support OpenClaw?

No. The Superpowers repository documents installers for fourteen agents — Claude Code, Codex CLI, Cursor, Gemini CLI, OpenCode, Pi and Hermes Agent among them — and OpenClaw is not on that list. There is no openclaw plugins install path for it.

Can I use the skills anyway?

Yes. Superpowers skills are plain SKILL.md documents under the repository's skills/ directory, and openclaw skills install accepts a git:<repo> ref or a local skill directory. The content ports; only the plugin wiring does not.

What does not carry over?

The plugin-level surface. Slash commands such as /superpowers:brainstorm belong to the Claude Code plugin system, and the automatic workflow hand-offs are enforced by the host agent rather than by the markdown. On OpenClaw you get the skills as documents the agent can load, not as a methodology it is compelled to follow.

Where do installed skills land?

In the agent workspace at ~/.openclaw/workspace/skills/<name>/SKILL.md. Add --global to install into the shared managed skills directory instead, which is what you want when several agents should see the same skill.

How do I check a skill is actually loadable?

Run openclaw skills check, which reports which skills are ready, visible, or missing requirements. openclaw skills list --eligible narrows the list to the ones that are genuinely ready to use.