# Superpowers for GitHub Copilot

This repository includes Superpowers skills. Use them before writing code, not after.

## Required Startup Workflow

Before responding to any task:

1. Load the core Superpowers skill:
   - First try `.superpowers/skills/using-superpowers/SKILL.md` (for symlink-based installs)
   - If missing, use `skills/using-superpowers/SKILL.md` (for in-repo usage)
2. Load Copilot tool mapping:
   - First try `.superpowers/skills/using-superpowers/references/copilot-tools.md`
   - If missing, use `skills/using-superpowers/references/copilot-tools.md`
3. If there is even a 1% chance a skill applies, load that skill's `SKILL.md` before continuing.

## Instruction Priority

1. User instructions and repository instructions (`AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, direct requests)
2. Superpowers skills
3. Default agent behavior

If instructions conflict, follow the user.

## Skill Invocation Rules

GitHub Copilot does not provide a dedicated Skill tool. Invoke skills by reading their `SKILL.md` files directly.

For rigid skills (for example, TDD and debugging), follow steps exactly. For flexible skills, adapt principles to context.

## Red Flags

If you think any of these, stop and load a skill first:

- "This is simple; I can skip skills."
- "I should gather context first."
- "I remember this skill already."
- "I can do one quick edit before checking skills."

## Workflow Priority

When multiple skills might apply:

1. Process skills first (`brainstorming`, `systematic-debugging`)
2. Implementation skills second (`writing-plans`, `executing-plans`, `subagent-driven-development`)

## Scope and Security

- Only access files relevant to the user request.
- Do not expose secrets, credentials, or sensitive values.
- Do not exfiltrate repository data to external services.
- Use least-privilege tools where possible.