Skills activate automatically
When you start a session with Superpowers installed, theusing-superpowers skill is discovered immediately. From that point, your agent checks for relevant skills before every response and every action. You don’t invoke skills yourself — the agent does.
The rule is strict: if there is even a 1% chance a skill applies to what the agent is doing, it must invoke that skill before responding. This is not optional behavior — it’s enforced by the
using-superpowers skill itself.Example prompts that trigger skills
Try any of these in a new session to see skills activate:| What you say | Skill triggered |
|---|---|
| ”Help me plan this feature” | brainstorming |
| ”Fix this bug” | systematic-debugging |
| ”Let’s start implementing” | test-driven-development |
| ”Review what I’ve built so far” | requesting-code-review |
A real example: building a feature
Here’s what a typical session looks like from the moment you describe something to build.You describe what you want to build
You say something like: “I want to add rate limiting to our API.”The agent recognizes this as a new feature request and invokes the
brainstorming skill before doing anything else.The agent asks questions one at a time
The
brainstorming skill instructs the agent to ask clarifying questions one at a time — never all at once. For example:“What kind of rate limiting are you thinking — per user, per IP, or per API key?”After you answer, it asks the next question. This continues until the agent has enough to propose a design.
The agent proposes 2–3 approaches with trade-offs
Once it understands the problem, the agent presents two or three concrete design options with trade-offs — not a single recommendation with assumed requirements. You choose the direction.
The agent presents the design in sections for your approval
Rather than dumping a wall of spec text, the agent presents the design in short, digestible sections and waits for your sign-off on each. When you approve, it saves the spec to a file.
The agent writes a detailed implementation plan
The
writing-plans skill produces a task list where every item is scoped to 2–5 minutes of work, includes exact file paths, complete code, and explicit TDD steps. No placeholders.Manually invoking a skill
Skills are designed to activate automatically, but you can also invoke them by name if you want to jump directly into a specific workflow.- Claude Code
- OpenCode
- Gemini CLI
- Cursor
- Codex
Use the Or the agent will invoke it via the
Skill tool. In your conversation, you can direct the agent:Skill tool directly when it detects the context matches.Verify the agent is following skills
After your first session, look for these signs that Superpowers is working:- The agent asks questions before writing code — brainstorming is active.
- The agent writes a failing test before any implementation — TDD is active.
- The agent announces which skill it’s using before beginning a workflow.
- The agent blocks on critical review findings rather than moving forward.
Next steps
The Workflow
Understand the full seven-step development loop from idea to merged branch.
Skills Reference
Explore every skill: TDD, debugging, subagent orchestration, code review, and more.