Using Superpowers (Skills System)

This skill teaches agents HOW to use the skills system. It's meta.

CRITICAL RULE

<MANDATORY> If there is even a 1% chance a skill might apply to what you are doing, you MUST check the index and load it.

This is NOT optional. This is NOT negotiable. You cannot rationalize your way out of this. </MANDATORY>

---

Part 1: Using Existing Skills

The Flow

User Request → Scan SKILL-INDEX.md → Match Skills → Load Relevant → Announce → Apply

Step-by-Step

1. Receive user request
2. Scan skills/SKILL-INDEX.md (lightweight, categorized)
3. Identify skills that might apply (by name, category, triggers)
4. Load the specific skill: skills/{skill-name}/SKILL.md
5. Announce: "Using [skill-name] for [purpose]"
6. Apply the skill's rules and patterns
7. Respond to user with skill-informed answer

Red Flags (You're Rationalizing)

Your Thought	The Reality
"This is just a simple question"	Simple questions have best practices too
"I already know this"	Skills evolve. Read the current version.
"Let me explore the code first"	Skills tell you HOW to explore
"This doesn't need a formal skill"	If a skill exists, use it
"The skill is overkill"	Discipline prevents mistakes
"I'll just do this one thing first"	Check skills BEFORE doing anything
"I know what the user means"	Skills provide the HOW, not just WHAT

Skill Priority

When multiple skills could apply:

1. Process skills first → debugging, planning, tdd, verification
2. Domain skills second → react, python, nextjs, stripe
3. Integration skills third → firebase, supabase, clerk

Example:

• "Fix this bug" → systematic-debugging first, then domain skills
• "Add Stripe" → stripe-integration + framework skill

Token Economy

• Index scan: ~500 tokens (ALWAYS acceptable)
• Full skill load: 1000-5000 tokens each
• Rule: Load max 2-3 skills per task
• Summarize skill content, don't quote entire files

---

Part 2: Detecting User Teaching

The User is Teaching When...

Signal	Example	Action
Explicit rule	"siempre usa X", "regla:", "nunca hagas Y"	Confirm & create
Correction	"no, hacelo asi..." + shows pattern	Ask if should persist
Preference	"prefiero X sobre Y porque..."	Note for skill
Repeated pattern	Same correction 3+ times	Propose skill creation
Best practice	"buena practica:", "el estandar es..."	Confirm & create
Style guide	"en este proyecto usamos..."	Add to project skill
Framework rule	"en React siempre...", "en Python nunca..."	Add to language skill

Detection Protocol

1. Recognize the teaching moment

   • User is explaining HOW, not just WHAT
   • User is correcting your output with a pattern
   • User says "siempre", "nunca", "regla", "preferencia"

2. Confirm intent

   • "¿Queres que guarde esto como skill/rule?"
   • "¿Es una preferencia tuya o un estandar del proyecto?"

3. Clarify scope

   • "¿Para que lenguaje/framework?" (python, react, general)
   • "¿Skill existente o nuevo?"
   • "¿Solo este proyecto o global?"

4. Preview the rule before creating

5. Create in skills/{skill-name}/rules/_custom-{name}.md

6. Rebuild index: bun run skills/_scripts/generate-index.ts

Rule File Format

---
title: Rule Title
impact: HIGH | MEDIUM | LOW
tags: tag1, tag2
source: user-taught
date: YYYY-MM-DD
---

## Rule Title

[Why this matters - one sentence]

**Do:**
```code
// correct pattern

Don't:

// incorrect pattern

[Optional: Reference or context]

### Scope Resolution

| User Says | Skill Location |
|-----------|----------------|
| "en Python siempre..." | `skills/python/rules/_custom-{name}.md` |
| "en React..." | `skills/react/rules/_custom-{name}.md` |
| "en este proyecto..." | `skills/general/rules/_custom-{name}.md` |
| "siempre que uses X..." | `skills/{x}/rules/_custom-{name}.md` |
| New domain entirely | Create new `skills/{name}/` folder |

---

## Part 3: Maintenance Commands

```bash
# After adding/modifying skills
bun run skills/_scripts/generate-index.ts

# To sync external sources (anthropics, vercel-labs)
bun run skills/_scripts/sync-external.ts

# To rebuild SKILL.md from rules/ folder
bun run skills/_scripts/build.ts

---

Quick Reference

Skill Categories (from SKILL-INDEX.md)

Category	Examples	When to Use
AI Agents & LLM	langgraph, crewai, rag-engineer	Building AI features
Development	react-patterns, nextjs, python	Coding specific stacks
Testing & QA	playwright, tdd-workflow	Testing strategies
Cybersecurity	pentest-*, ethical-hacking	Security audits
Integrations	stripe, firebase, supabase	Third-party APIs
Marketing	seo-, cro-, copywriting	Growth features
Infrastructure	docker, aws-, git-	DevOps tasks
Workflow	writing-plans, executing-plans	Process management

Skill Types

Rigid (TDD, debugging, verification): Follow EXACTLY. Don't adapt away discipline.

Flexible (patterns, integrations): Adapt principles to context.

The skill itself tells you which type it is.

---

The Golden Rule

Instructions say WHAT to do. Skills say HOW to do it well.

"Add Stripe payments" is WHAT. stripe-integration skill tells you HOW.

Never skip the HOW.