Superpowers

Superpowers is a software-development methodology packaged as 15 composable agent skills. It pushes the agent through brainstorming, a written plan, isolated worktrees, subagent-driven implementation, test-driven development, code review, and branch cleanup instead of jumping straight to code.

The skills are the product. A small SessionStart hook exists only to make the agent load the using-superpowers bootstrap before it starts work.

Source and revision

What is actually in the repository

The 15 skills are grouped by phase, not by technology:

The workflow the README describes is a chain: brainstorm the intent, write a plan with exact file paths and verification steps, create an isolated worktree, implement task by task, review each task, then merge or discard the branch. TDD is enforced as RED-GREEN-REFACTOR; the skill says code written before a failing test gets deleted.

That is opinionated. It is closest to a team process encoded as skills, not a generic prompt pack.

Install

/plugin install superpowers@claude-plugins-official

/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

The session hook and telemetry

The plugin registers hooks/session-start, a Bash script that reads skills/using-superpowers/SKILL.md and injects it into the session context. That is how the agent knows to check for a relevant skill before acting. The hook itself makes no network call.

There is one network path to know about: the optional visual companion in brainstorming loads a Prime Radiant logo from the project website. The request includes the Superpowers version, not the project, prompt, or agent. It is on by default and can be disabled:

export SUPERPOWERS_DISABLE_TELEMETRY=1

The README says Superpowers also honors Claude Code's DISABLE_TELEMETRY and CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC settings. If you block the hook, the skills still exist but may not auto-trigger.

Permissions and footprint

The skills can direct the agent to use Bash, Edit, and worktree commands. That is the point of the framework, and it is also the boundary: Superpowers is not a sandbox. Review the individual skill that touches your repository before granting it a broad tool policy.

Licensing and commercial support

MIT. Prime Radiant offers commercial support for enterprises, but the skills are not paywalled and no account is required.

Limits

• Opinionated by design. TDD-first and spec-first. Teams with a different process will fight the chain.
• The hook matters. Without SessionStart, automatic skill discovery is weaker; the harness must support the hook.
• Not a sandbox. The skills instruct the agent to run commands and edit files. The host's permission model is the real control.
• Per-harness behavior differs. Install paths and hook support vary across the 16+ documented harnesses.
• No independent benchmark here. This page did not measure the claimed speed or token improvements from Superpowers 6.
• Adoption is not evidence. Star counts are large but say nothing about fit for your repository.
• Not runtime-tested for this review. The hook and skills were read, not executed.

Verify it yourself

1. Clone at 2ce8aa1 and read skills/using-superpowers/SKILL.md, brainstorming, and test-driven-development before installing.
2. Read hooks/session-start. Confirm it only reads the local skill file and injects context.
3. Install in a scratch repository and start a new session. Ask for a small feature; the agent should ask clarifying questions before writing code.
4. Check whether the hook ran and whether the using-superpowers text appears in context.
5. If you care about the logo request, set SUPERPOWERS_DISABLE_TELEMETRY=1 and confirm the request stops.
6. Remove the plugin through the harness's plugin manager. The repository documents no separate uninstaller; removal steps vary by harness.

Review transparency

This is an original guide with short attributed references. It does not reproduce the full skill text.

Related

• Superpowers 6: Parallel Agents — the orchestration model behind the newer workflow
• Skills Collection Overview — the rest of the catalog
• How to Vet an Agent Skill — review a multi-skill framework before installing it
• Building Antigravity + OpenCode Skills — first-person build log