chidekina/aria-superpowers/token-saver

Claude Code and agent skill by chidekina/aria-superpowers

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Token optimization guide for Claude Code. This skill should be used when the user wants to reduce API costs, learn model routing strategies, or understand how to use free/cheaper tools (Ollama, Haiku) instead of burning expensive Claude tokens on routine tasks. Triggers on: save tokens, reduce costs, token tips, cheap claude, model routing, ollama setup, token optimization.

Install

npx vskill@latest install chidekina/aria-superpowers/token-saver

SKILL.md

Token Saver — Claude Code Cost Optimization

Overview

A routing and habits guide to cut Claude API costs. The core idea: match the tool to the task — use free/cheap tools for simple work, and reserve expensive models for genuinely hard problems.

Quick Reference

FREE    → Ollama  (commits, summaries, translations, explanations)
CHEAP   → Haiku   (quick lookups, simple one-liners)
DEFAULT → Sonnet  (daily coding, features, bugs)
DEEP    → Opus    (architecture, security, complex refactors only)

Tier 0 — Ollama (FREE, runs locally)

Install once, zero API cost forever.

Setup:

bash
curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2:3b

Always route these to Ollama:

TaskCommand
Commit messageollama run llama3.2:3b "write a conventional commit for: <diff>"
Summarize textollama run llama3.2:3b "summarize: <text>"
Translate PT↔ENollama run llama3.2:3b "translate to english: <text>"
Explain a functionollama run llama3.2:3b "explain this code: <snippet>"
PR descriptionollama run llama3.2:3b "write a PR description for: <summary>"
Classify textollama run llama3.2:3b "classify as bug/feature/chore: <text>"

Rule: No deep reasoning needed + short output = Ollama.

Tier 1 — Haiku (Cheapest Claude)

bash
/model claude-haiku-4-5-20251001

Good for: Simple lookups, yes/no questions, short edits with clear instructions, quick one-liners.

Tier 2 — Sonnet (Default)

bash
/model claude-sonnet-4-6

Good for: Feature implementation, bug fixes, code reviews, most daily coding tasks. Stay here unless stuck.

Tier 3 — Opus (Use Sparingly)

bash
/model claude-opus-4-6

Good for: System architecture, security audits, complex multi-system refactors. Not for: Anything Sonnet handles on the first try.

Read Code Surgically

Every line loaded into context costs tokens.

Instead ofDo this
"Read this file and find the auth logic""Find the authenticate function in auth.service.ts"
"Look at the whole component""Show me just the useEffect hooks in Dashboard.tsx"
Opening a 500-line fileGrep for the symbol first, then read only that block

Scope Requests Tightly

Bundling tasks inflates both input and output tokens.

Instead of: "Fix this bug, refactor the function, add tests, and update the docs" Do: Fix the bug → commit → then ask for the next thing separately.

Stop Retry Loops Early

Re-running the same failing prompt 5 times multiplies cost with no gain.

  • Stop after 2 failed attempts
  • Rethink the approach or ask a more specific question
  • A different angle costs fewer tokens than brute-forcing the same one

Biggest Wins (in order)

  1. Move commits/translations/summaries to Ollama — these are 100% free
  2. Stay on Sonnet — don't upgrade to Opus by default
  3. Read functions, not files — surgical reads cut context size dramatically
  4. One task per prompt — smaller prompts, smaller responses

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

See the full security report

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

See evaluation results

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/chidekina/aria-superpowers/token-saver/badge)](https://verified-skill.com/skills/chidekina/aria-superpowers/token-saver)

Something wrong with this skill or its page? Tell us.