Brainstorming Skill

Brainstorming is the first skill in the Superpowers workflow (skill source). It activates before any creative work, such as a new feature, component or behavior change, and turns a rough idea into an approved design through dialogue. Nothing gets implemented until you approve the design the skill calls for.

The current version sorts every request into one of three paths and says the classification out loud so you can override it.

Key Features

• Shared understanding first: the agent identifies the intended outcome, who it is for and what success looks like, then writes that back for you to correct.
• Three paths: a spike answers a feasibility question with throwaway work; a bounded change to existing code gets a short design in chat; architectural work gets clarifying questions, 2 to 3 approaches with trade-offs, a sectioned design, and a written spec.
• One-way ratchet: when in doubt, take the heavier path; hidden complexity upgrades the path mid-task, never the reverse.
• Hard gate: no code, scaffolding or dependency installs before approval. Read-only exploration is allowed.
• Spec output: architectural designs are saved to docs/superpowers/specs/YYYY-MM-DD-<topic>-design.md, self-reviewed, then handed to Writing Plans.
• Visual companion: an optional browser view, offered only when a question is clearer shown than described.

Use Cases

• Starting a new project or subsystem.
• A feature request where the real goal is unclear.
• A quick "can we even do X?" spike.

Pricing

Free and open source under the MIT license. Prime Radiant, the company behind Superpowers, sells commercial support to enterprises. The practical cost is tokens: process skills add questions, reviews and subagent runs.

Getting Started

Superpowers installs as one plugin, so this skill arrives with the rest of the library (15 skills in v6.4.1). In Claude Code run /plugin install superpowers@claude-plugins-official. Cursor uses /add-plugin superpowers, Gemini CLI uses gemini extensions install https://github.com/obra/superpowers, and the README lists commands for Codex, GitHub Copilot CLI, OpenCode, Pi and other harnesses. Skills trigger on their own; you can also ask for one by name. Then describe what you want to build and answer the questions one at a time.

Limitation: the approval gates add round-trips, which feels slow for small tasks. The visual companion loads the Prime Radiant logo, which reports the Superpowers version; set SUPERPOWERS_DISABLE_TELEMETRY to turn that off.

FAQ

Can I skip straight to code?

Only by telling the agent explicitly; the skill treats "too simple to need a design" as a red flag.

Where does the spec go?

Into your repo, under docs/superpowers/specs/, and it is committed.

Alternatives

• Spec Kit: GitHub's spec-driven development toolkit.
• Planning with Files: persistent markdown planning.
• Doc Co-Authoring: Anthropic's example skill for writing documents with the agent.

Conclusion

It stops agents from building the wrong thing fast. Accept the extra questions on big work. More in the skills hub.