convez.ai
Skill
linkinwardi

Token_Optimizer_SKILL

linkinwardi/Token_Optimizer_SKILL

Ver no GitHub
InicialPluginScore · de 0 a 100

Optimize Cloud IA tokens for Vibe Coding using this automation skill. Maximize efficiency and cut Claude API costs. Perfect for developers building robust apps, tools, or even free games online. It includes PowerShell scripts for quick setup and strict ignore rules to prevent context bloat, saving quota and minimizing expensive token waste.

Instalação

$npx convez-ai linkinwardi/Token_Optimizer_SKILL

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

vibe-coding-skills

Skills for token-efficient agentic coding on Claude Code and the Claude Platform.

Currently ships one skill:

vibe-coding-token-optimizer

Cuts input-token spend and quota drain during long "vibe coding" sessions by setting up three mechanisms that actually take effect in Claude Code:

  • CLAUDE.md — the auto-loaded rules file (plan-first, compressed output, a "do not read" directory list), kept static so it stays prompt-cached.
  • .claude/settings.json → permissions.deny — Read(...) rules that keep heavy dirs (node_modules/, dist/, build/, bin/, obj/, coverage/, *.log) out of auto-context.
  • .gitignore alignment — so Claude Code's file picker skips build output.

A bare .claudeignore is deliberately not used — stock Claude Code does not honor it. For hard enforcement, see the skill's references/pretooluse-hook.md.

Includes an idempotent PowerShell deploy script (deploy.ps1) that creates and merges these files for you (Windows / cross-platform pwsh).


Install

Option A — Plugin marketplace (recommended)

Inside a Claude Code session:

/plugin marketplace add linkinwardi/Token_Optimizer_SKILL
/plugin install vibe-coding-token-optimizer@vibe-coding-skills

Restart the session (or /reload-plugins). The skill loads automatically and triggers when you talk about token usage, Claude Code cost, context size, etc.

Option B — Manual (copy the skill folder)

For all your projects (personal scope):

git clone https://github.com/linkinwardi/Token_Optimizer_SKILL.git
cp -r vibe-coding-skills/plugins/vibe-coding-token-optimizer/skills/vibe-coding-token-optimizer \
      ~/.claude/skills/

For one repo only (project scope, shared with anyone who clones that repo): copy the same folder into <repo>/.claude/skills/ instead. Start a new session.

Option C — Claude.ai / Claude API

Upload the packaged .skill file via Settings → Capabilities → Skills (paid plans), or the Skills API. Build the .skill from this repo with Anthropic's skill packager, or download it from this repo's Releases.


Use

Once installed, just ask in natural language — no slash command needed:

  • "My Claude Code sessions are burning too many tokens, optimize this project."
  • "Set up a CLAUDE.md and stop the agent from reading node_modules."
  • "Deploy the token optimizer for a .NET repo."

Or run the automation directly from a project root:

pwsh ./deploy.ps1                  # create CLAUDE.md + settings + gitignore
pwsh ./deploy.ps1 -Stack dotnet    # tune deny list for a stack
pwsh ./deploy.ps1 -PruneLegacy     # also remove *_legacy.md (asks first)

License

MIT — see LICENSE.

Ver o original no GitHub

Bom para

Relacionados