seed-prompt-cache-anchor

Restructure prompts to maximize Anthropic cache hits with stable prefixes and breakpoints.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/JBODE-mhhs/Zeus2.0-public --skill seed-prompt-cache-anchor
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: seed-prompt-cache-anchor
Source: https://github.com/JBODE-mhhs/Zeus2.0-public/tree/main/community/seeds/seed-prompt-cache-anchor
Command: npx skills add https://github.com/JBODE-mhhs/Zeus2.0-public --skill seed-prompt-cache-anchor

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill restructures prompts to maximize Anthropic prompt caching by moving the stable instructions to the top, marking the cache breakpoint correctly, and removing per-turn-changing junk from above the breakpoint.

Core Features & Use Cases

  • Identify the stable prefix in prompts that should not vary between calls.
  • Place a clear cache breakpoint to separate stable content from per-turn content.
  • Verify cache hit rates and the cache_creation/cache_read metrics to minimize costs.
  • Applicable to long-running agent workflows using Anthropic models to reduce per-call costs.

Quick Start

Place the stable, per-turn-invariant instructions at the top and clearly demarcate the breakpoint so that only changing content sits after it.

Frequently Asked Questions about seed-prompt-cache-anchor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I maximize Anthropic prompt caching hits for long-running agent workflows?

Maximize Anthropic prompt caching by restructuring prompts to move stable instructions to the top and marking the cache breakpoint. This ensures only per-turn-changing content sits after the breakpoint, increasing cache hits for long-running agent workflows.

What is a stable prefix and why does it matter for prompt caching?

A stable prefix is the per-turn-invariant instruction block in your prompt. Keeping it above the cache breakpoint allows Anthropic models to reuse cached tokens, minimizing per-call costs in long-running agent workflows.

How do I verify cache hit rates using cache_creation and cache_read metrics?

Verify cache hit rates by auditing the cache_creation and cache_read metrics returned by the Anthropic API. This confirms your stable prefix placement and breakpoint demarcation are successfully generating expected cache hits.

Does prompt caching work with prompts that change every turn?

Yes, prompt caching works by isolating stable content from per-turn-changing content. You must remove variable junk from above the breakpoint so only the stable prefix is cached, while dynamic content sits below it.

What is the best way to structure prompts to reduce Anthropic API costs?

The best way to reduce Anthropic API costs is to restructure prompts by placing stable instructions at the top, marking a clear cache breakpoint, and verifying cache usage metrics to prevent per-turn-changing junk from invalidating the cached prefix.

Why does my Anthropic prompt cache miss when I add new instructions?

Your prompt cache misses because adding new instructions above the cache breakpoint alters the stable prefix. You must keep per-turn-changing junk below the breakpoint to maintain cache hits and minimize per-call costs.