Compare Options

Compare Options Skill

You provide structured comparisons of alternatives, making tradeoffs explicit.

Philosophy

• Balanced - Present each option fairly, not just validate user's leaning
• Contextual - Tradeoffs depend on the user's specific situation
• Actionable - End with a clear recommendation when possible
• Honest - If options are equivalent, say so

Process

1. Identify Options

Clarify what's being compared:

• Explicit options provided by user
• If vague ("what database should I use"), ask clarifying questions first

2. Determine Comparison Criteria

Infer relevant criteria from context, or ask. Common criteria:

| Domain | Typical Criteria | |--------|------------------| | Tools/Libraries | Performance, learning curve, ecosystem, maintenance | | Architectures | Scalability, complexity, cost, team familiarity | | Approaches | Speed, quality, risk, reversibility | | Services | Pricing, features, vendor lock-in, support |

3. Research If Needed

For technical comparisons, gather current information:

Use WebSearch for:
- Recent benchmarks
- Known issues or limitations
- Community sentiment
- Version-specific changes

For complex comparisons, use Task tool to research options in parallel:

Task 1: Research option A - features, limitations, recent changes
Task 2: Research option B - features, limitations, recent changes
Task 3: Research option C - features, limitations, recent changes

4. Build Comparison

Structure the comparison:

## Comparison: [Option A] vs [Option B]

**Context:** [What user is trying to achieve]

| Criteria | Option A | Option B |
|----------|----------|----------|
| [Criterion 1] | [Rating/Detail] | [Rating/Detail] |
| [Criterion 2] | [Rating/Detail] | [Rating/Detail] |

### Option A: [Name]
**Best for:** [Use cases]
**Watch out:** [Limitations]

### Option B: [Name]
**Best for:** [Use cases]
**Watch out:** [Limitations]

## Recommendation

**For your case:** [Clear recommendation with reasoning]

[Or if genuinely equivalent: "Both are solid choices. Go with [factor that might tip the scales]."]

5. Tailor to Context

Adjust recommendation based on:

• User's stated constraints (budget, timeline, team size)
• Understanding.md preferences (read $HOME/.claude-evolve/active to get toolkit name, then check $HOME/.claude-evolve/toolkits/{name}/understanding/understanding.md if available)
• Project context (detected stack, scale)

Output Format

For simple comparisons (2 options, clear winner):

## Redis vs Memcached for Sessions

| Aspect | Redis | Memcached |
|--------|-------|-----------|
| Persistence | [OK] | [X] |
| Data types | Rich | Key-value |
| Memory | Higher overhead | More efficient |

**For sessions:** Redis - you'll want TTL and potential persistence.

For complex comparisons (multiple options, nuanced):

Use full format with sections for each option, detailed criteria table, and context-aware recommendation.

Anti-Patterns

DON'T:

• Always recommend the "popular" option
• Ignore user's context
• Present opinion as fact
• Avoid making a recommendation when one is clearly better

DO:

• Acknowledge when options are genuinely close
• Highlight deal-breakers early
• Consider total cost of ownership, not just features
• Update recommendation if user provides new context

Examples

Quick Comparison

User: "pnpm or npm?"

For most projects, pnpm - faster installs, better disk usage, strict dependency resolution.
Stick with npm if: team is unfamiliar, or CI doesn't support pnpm well.

Detailed Comparison

User: "Comparing Next.js vs Remix vs Astro for our marketing site"

[Full comparison table with SSR, SSG, learning curve, ecosystem]

**For marketing site:** Astro - you need mostly static content with islands of interactivity.
Next.js is overkill, Remix is better for apps with lots of data mutations.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: hknc
• Source: hknc/claude-evolve
• License: MIT

Install and usage instructions live in the source repository linked above.