LLM Cost Optimizer

LLM Cost Optimizer

What this skill does

This skill audits an LLM application's prompts, call patterns, and model selection to identify cost reduction opportunities. It covers prompt caching, model routing (right-sizing), token reduction, batching, and output length control — the techniques that typically cut LLM costs by 40–80% without sacrificing quality.

How to use

Claude Code / Cline

Copy this file to .agents/skills/llm-cost-optimizer/SKILL.md in your project root.

Then ask:

• "Use the LLM Cost Optimizer to audit our AI application."
• "How can I reduce our OpenAI API costs? Here are our prompts..."

Provide:

• Your system prompt(s)
• Approximate daily call volume
• Which model(s) you're using
• Typical input/output token counts if known
• Whether calls are real-time (low latency required) or batch (latency tolerant)

Cursor / Codex

Paste your prompts, call patterns, and current monthly spend alongside these instructions.

The Prompt / Instructions for the Agent

When asked to optimize LLM costs, audit the following areas in order of typical savings impact:

Audit 1 — Prompt Caching (savings: 50–90% on repeated prefixes)

Check: Does the system prompt stay the same across calls?

If yes, enable prompt caching. The system prompt is sent once and cached — subsequent calls only pay for the new user tokens.

# Anthropic Claude — cache_control on system prompt
response = client.messages.create(
    model="claude-opus-4-6",
    system=[{
        "type": "text",
        "text": your_system_prompt,
        "cache_control": {"type": "ephemeral"}  # cached for 5 minutes
    }],
    messages=[{"role": "user", "content": user_message}]
)

# OpenAI — automatic prompt caching for prompts > 1024 tokens
# No code change needed — cached automatically, check usage.prompt_tokens_details.cached_tokens

When it applies: Any app where the system prompt is > 1024 tokens and reused across calls. Support bots, coding assistants, document analyzers.