---
name: using-superpowers
description: Bootstrap skill for the pi-superpowers bundle. Always relevant — read at the start of every session and before responding to any non-trivial request, to decide whether another skill in this bundle applies.
---

# Using Superpowers

You have access to the `pi-superpowers` skill bundle. Treat it as process discipline, not optional flair.

## The rule

Before answering any non-trivial request, scan the loaded skill descriptions. If there is even a small chance one applies, load and follow it. Announce briefly which skill you are using and why.

If a skill clearly does not fit, proceed without it — but make that a deliberate decision, not a default.

## Priority order when multiple skills could apply

1. **Process skills first** — `brainstorming`, `systematic-debugging`, `writing-plans`. These determine *how* to approach the task.
2. **Execution skills second** — `test-driven-development`, `using-git-worktrees`, `dispatching-parallel-agents`.
3. **Closing skills last** — `verification-before-completion` before you claim done.

## Common triggers

| User says / situation | Skill to consider |
|---|---|
| "Let's build / add / design X" with fuzzy requirements | `brainstorming` |
| Multi-file or multi-step change | `writing-plans` |
| "It's broken", "X isn't working", unexpected output | `systematic-debugging` |
| New behavior with a testable contract | `test-driven-development` |
| Risky/experimental change in a shared repo | `using-git-worktrees` |
| 2+ independent tasks queued up | `dispatching-parallel-agents` |
| About to say "done" / "fixed" / "shipped" | `verification-before-completion` |

## User instructions always win

If the user's CLAUDE.md, AGENTS.md, or a direct request contradicts a skill, follow the user. Skills are defaults, not laws.