From magic-powers
Use when deciding which model or agent to use for a task - guides cost-optimized model selection based on task complexity
npx claudepluginhub kienbui1995/magic-powers --plugin magic-powersThis skill uses the workspace's default tool permissions.
Not every task needs the most expensive model. Route tasks to the right model based on complexity, saving cost without sacrificing quality.
Generates design tokens/docs from CSS/Tailwind/styled-components codebases, audits visual consistency across 10 dimensions, detects AI slop in UI.
Records polished WebM UI demo videos of web apps using Playwright with cursor overlay, natural pacing, and three-phase scripting. Activates for demo, walkthrough, screen recording, or tutorial requests.
Delivers idiomatic Kotlin patterns for null safety, immutability, sealed classes, coroutines, Flows, extensions, DSL builders, and Gradle DSL. Use when writing, reviewing, refactoring, or designing Kotlin code.
Not every task needs the most expensive model. Route tasks to the right model based on complexity, saving cost without sacrificing quality.
| Model | Agent | Cost | Best For |
|---|---|---|---|
| Opus | architect | $$$$$ | Complex architecture, novel design, ambiguous requirements |
| Sonnet | debugger, ui-designer, (default) | $$ | Implementation, debugging, most coding tasks |
| Haiku | reviewer | $ | Code review, verification, simple lookups, formatting |
Cost ratios (approximate):
WHEN receiving a task:
1. Can Haiku handle this? (review, verify, format, lookup)
→ YES: Use reviewer agent or --model haiku
2. Is this standard implementation/debugging?
→ YES: Use default Sonnet (or debugger/ui-designer agent)
3. Does this require deep reasoning about novel problems?
→ YES: Use architect agent (Opus)
| Pattern | Problem | Fix |
|---|---|---|
| Using Opus for everything | 25x cost for same result | Default to Sonnet |
| Using Opus for code review | Overkill, slow | Use Haiku reviewer |
| Using Haiku for implementation | Too many mistakes | Use Sonnet |
| Not using agents at all | Missing cost savings | Dispatch to right agent |
# Quick review with Haiku
claude -p "review the last commit" --model haiku
# Debug with Sonnet (default)
claude -p "fix the failing test in auth.py"
# Architecture with Opus
claude -p "design the caching layer" --model opus
# Budget cap
claude --max-budget-usd 0.50 -p "implement the login form"
Track your usage in Claude Code settings or Vertex AI console: