Superpowers Skills for Claude Code: 14 Listed Skills Ranked

Mert DüzgünPublished November 15, 20267 minute read

Superpowers is a plugin by Jesse Vincent and Prime Radiant that adds a full software development workflow to Claude Code (and other coding agents) as a set of skills. We list 14 skills from the obra/superpowers repo on CCM, and the top one, brainstorming, has about 423,000 skills.sh installs. The repo has about 315,000 GitHub stars (both as of November 13, 2026).

What Superpowers is

The README calls Superpowers "a complete software development methodology for your coding agents." It is a folder of skills plus one bootstrap skill, using-superpowers, that tells Claude to check for a matching skill before it does anything else.

The using-superpowers skill says that if there is "even a 1% chance a skill might apply," Claude must invoke it, and it must do so before any response, including clarifying questions or a quick look at the codebase. The skill gives two examples: "Let's build X" starts with brainstorming, and "Fix this bug" starts with systematic-debugging.

Jesse Vincent announced the first release in his post on October 9, 2025. The README credits him and the team at Prime Radiant.

How to install Superpowers in Claude Code

The plugin route gives you every skill plus the bootstrap. The /plugin commands below are typed inside Claude Code, not in your shell.

From Anthropic's official plugin marketplace:

Inside Claude Code, type:

/plugin install superpowers@claude-plugins-official

Or from the Superpowers marketplace, which also carries the related plugins covered below:

Inside Claude Code, add the marketplace:

/plugin marketplace add obra/superpowers-marketplace

Inside Claude Code, install the plugin:

/plugin install superpowers@superpowers-marketplace

CCM lists the Superpowers marketplace plugin (version 7.0.2) and a copy in the repo's own marketplace. To use the repo's own marketplace:

Inside Claude Code, add the marketplace:

/plugin marketplace add obra/superpowers

Inside Claude Code, install the plugin:

/plugin install superpowers@superpowers-dev

For one or two skills, use skills.sh. This is the command CCM shows on each skill page:

In your terminal, install brainstorming (replace brainstorming with any skill name from the table below):
```
npx -y skills add obra/superpowers --skill brainstorming --agent claude-code
```

Single skills installed this way come without using-superpowers. They still trigger from their own descriptions, but Claude is no longer told to check for a skill before every response.

The Superpowers workflow, step by step

The README lays out a basic workflow, and each skill hands off to the next.

Before anything gets built, brainstorming runs. Claude reads the project and any files you pointed at, plays your request back in its own words, and asks one open question at a time about what you want and why. It lays out approaches for you to pick from, then writes a design doc (short for small changes, full for projects) and waits for your approval.
After you approve the design, using-git-worktrees sets up an isolated workspace on a new branch, runs project setup and checks that tests pass before work starts.
writing-plans then turns the approved design into a plan "for an engineer who has not seen this codebase or this spec." Each task names the exact files, the interfaces and the tests that prove it, broken into small steps such as "write the failing test" and "run it to make sure it fails."
To run the plan, pick subagent-driven-development or executing-plans. The subagent route sends each task to a fresh implementer subagent, then a reviewer checks spec compliance and code quality before the next task starts, with a whole-branch review at the end. The inline route has Claude do every task itself in one context, with one fresh review at the end. The README calls the first "most thorough" and the second "cheapest."
test-driven-development enforces one rule during implementation: "no production code without a failing test first." Code written before its test gets deleted.
requesting-code-review runs between tasks and reports issues by severity. Critical issues block progress.
Last, finishing-a-development-branch checks the tests, offers to merge, open a pull request, keep or discard the branch, and cleans up the worktree.

Two skills sit beside this path. systematic-debugging takes over when something breaks and forbids fixes until a root cause is found, working through four phases that start with reading the error, reproducing it and checking recent changes. verification-before-completion stops Claude from saying "done" or "tests pass" until it has run the command in that same message and read the output.

The README says it is "not uncommon" for the agent to "work autonomously for a couple hours at a time without deviating from the plan you put together."

The 14 Superpowers skills on CCM, ranked by installs

Counts are skills.sh installs as of November 13, 2026. Each name links to its CCM page.

Rank	Skill	What it does	skills.sh installs
1	brainstorming	Asks what you want and why, then writes a design doc before any code	423,255
2	systematic-debugging	Four-phase root cause process before any fix	326,655
3	writing-plans	Turns a spec into small tasks with exact files and tests	298,629
4	using-superpowers	Bootstrap that makes Claude check for skills before every response	283,664
5	requesting-code-review	Reviews work against the plan before you merge	276,536
6	test-driven-development	Red, green, refactor, with a failing test first every time	274,814
7	verification-before-completion	No success claim without fresh command output	270,303
8	executing-plans	Runs a plan inline in one session, one review at the end	257,544
9	subagent-driven-development	One fresh subagent and one review per task	251,990
10	receiving-code-review	Checks review feedback on its merits before acting on it	235,064
11	dispatching-parallel-agents	One agent per independent problem, run at the same time	231,302
12	writing-skills	Creates and tests new skills	231,263
13	using-git-worktrees	Sets up an isolated branch and workspace	230,229
14	finishing-a-development-branch	Merge, pull request, keep or discard, then cleanup	227,674

The 14th skill has more than half the installs of the first. The README also lists a diagnosing-superpowers skill that reads a session transcript when skills misbehave. We do not list it on CCM yet, so it has no count here.

On CCM, brainstorming also leads install copies by a wide margin: 1,046 in the 90 days to November 13, 2026, against 159 for using-superpowers in second place.

Other plugins in the Superpowers marketplace

These install the same way. Versions are as CCM lists them on November 14, 2026.

Inside Claude Code, add the marketplace (skip this if you added it above):
```
/plugin marketplace add obra/superpowers-marketplace
```

Plugin	What it does
episodic-memory (1.6.0)	Semantic search over past Claude Code and Codex conversations
superpowers-chrome (3.0.5)	Chrome DevTools Protocol access through a browsing skill or one MCP tool
elements-of-style (1.0.0)	Writing rules from Strunk's The Elements of Style (1918)
superpowers-lab (0.5.0)	Experimental skills: tmux for interactive CLIs, MCP server discovery, duplicate function detection
double-shot-latte (1.2.0)	Decides on its own whether Claude should keep working instead of asking you
claude-session-driver (4.0.1)	Starts and monitors other Claude Code sessions as workers through tmux
private-journal-mcp (2.0.1)	A private journal MCP server with local embeddings and search
superpowers-developing-for-claude-code (0.3.1)	Skills and docs for building Claude Code plugins, skills and MCP servers

Install any of them, inside Claude Code:

episodic-memory:

/plugin install episodic-memory@superpowers-marketplace

superpowers-chrome:

/plugin install superpowers-chrome@superpowers-marketplace

elements-of-style:

/plugin install elements-of-style@superpowers-marketplace

superpowers-lab:

/plugin install superpowers-lab@superpowers-marketplace

double-shot-latte:

/plugin install double-shot-latte@superpowers-marketplace

claude-session-driver:

/plugin install claude-session-driver@superpowers-marketplace

private-journal-mcp:

/plugin install private-journal-mcp@superpowers-marketplace

superpowers-developing-for-claude-code:

/plugin install superpowers-developing-for-claude-code@superpowers-marketplace

The marketplace also has superpowers-dev, the development branch. Its own description says you must uninstall other versions of Superpowers before you install it, so leave it alone unless you want to test unreleased changes.

The trade-off: tokens and time

The bootstrap is strict by design. using-superpowers asks Claude to invoke skills before any response, and its red-flags table rules out shortcuts such as "this is just a simple question." Each loaded skill adds text to the context, so a short question can cost a skill check first. The plan runners have a cost too: subagent-driven-development "costs a fresh context per task and per review," per its own SKILL.md. The skill tells Claude to use the least powerful model that can handle each role to keep that down.

The README's "When Something Goes Wrong" section names sessions where a skill fires when it should not or the agent "burns more tokens than you'd expect," and points you to the diagnosing-superpowers skill to find out why. If a full workflow is too much for a small fix, the inline executing-plans route is the cheaper one, and you can tell Claude to skip the workflow: using-superpowers says your instructions take precedence over skills.

Where to start

Install the plugin, then start a session with a feature request and let brainstorming run. It comes first in the workflow, and writing-plans, the next big step, builds on the design doc it writes.

If you want to try one skill first, take brainstorming or systematic-debugging. using-superpowers calls them the plugin's most common process skills. The other Superpowers plugins are in the marketplace table above.