Analyze

README

# analyze Multi-perspective analysis for any topic, decision, or piece of code. Use before making significant changes or when evaluating trade-offs.

SKILL.md

---
name: analyze
description: Multi-perspective analysis for any topic, decision, or piece of code. Use before making significant changes or when evaluating trade-offs.
---

# Analyze

## Description
Examine any problem from multiple perspectives before acting. Prevents tunnel vision and catches blind spots that a single-perspective analysis would miss.

## When to use
When the user asks "what do you think about X", "analyze this", "should I do X or Y", or when the agent detects a high-risk change (large refactor, dependency swap, architecture decision, API redesign).

## Instructions
Given a topic, file, decision, or problem, analyze from these perspectives:

### Technical perspective
- What's the current architecture and how does this fit?
- What are the trade-offs of each approach?
- What technical debt exists in this area?

### Risk perspective
- What could break? What's the blast radius?
- What's the rollback plan?
- Is there a safer incremental path?

### Maintenance perspective
- How hard is this to change or extend later?
- Will new team members understand it?
- Is it consistent with the rest of the codebase?

### Performance perspective
- Are there bottlenecks or N+1 problems?
- What's the scalability ceiling?
- Can this be cached, parallelized, or lazy-loaded?

### Security perspective
- Are there new attack surface areas?
- Is sensitive data handled correctly?
- Are dependencies introducing risk?

## Output format
Keep it concise:
- **Summary**: 1-2 sentences on the key finding
- **Per-perspective**: Top insight from each perspective
- **Recommendation**: Clear action with rationale

## Examples

Input: "Should we migrate from REST to GraphQL?"
Output:
- Technical: Reduces over-fetching but adds query complexity and resolver N+1 problems
- Risk: High-touch migration with no easy rollback once schema is consumer-facing
- Maintenance: Team needs GraphQL training; schema becomes the contract
- Performance: N+1 shifts from client to resolver layer — requires DataLoader
- Recommendation: Use GraphQL for new features only, keep REST for existing endpoints