OnlyTerp

OnlyTerp / prompt-cache-skills

Public

Drop-in prompt-caching fixes for the LLM agent harness you use. Point your AI coding agent at this repo and it ships the patches.

agents-md ai-skills aider anthropic claude-code
16
2
89% credibility
Found May 28, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A collection of drop-in fixes for AI coding assistant tools that have bugs preventing prompt caching from working, potentially saving users up to 10x on their AI API costs.

How It Works

1
💸 Your AI bills are too high

You're using an AI coding assistant and notice your API bills are much higher than expected

2
🔍 You find the right fix

You discover this repo has ready-made fixes for the exact tool you're using

3
🤖 Your AI reads the instructions

You simply ask your AI assistant to read this repo and apply the relevant fixes

4
Choose how to use it
💬

Tell your AI assistant to read the repo and apply matching fixes

📁

Install the skill bundle in your AI assistant's skills folder

5
⚙️ Fixes are applied automatically

Your AI checks each fix applies to your setup, makes the changes, and runs tests to confirm everything works

6
✅ Everything is verified

The tools verify your caching is now working by checking your actual API responses

🎉 You save money every day

Your AI assistant now uses caching properly, cutting your API costs by up to 90%

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is prompt-cache-skills?

This is a Python project that patches broken prompt-caching in popular AI coding tools. When your coding agent sends requests, caching lets the API remember previous context so you pay 10x less. But most harnesses have bugs: they cache the wrong parts, leave caching off by default, or don't implement it for certain providers. This repo gives you 13 drop-in fixes that your agent can apply automatically. You point your agent at the repo, ask it to apply relevant skills, and it reads each fix, checks if it applies to your setup, applies the patch, and verifies the fix worked. No manual research needed.

Why is it gaining traction?

The hook is the math. A 50-turn coding session might cost $7.50 when it could cost $0.75. The repo audits 13 different harnesses, documents exactly which ones are broken and why, and gives you a verification script to confirm caching works before and after each fix. You don't have to understand prompt-caching internals. You just point your agent at the repo and let it do the work.

Who should use this?

If you use Aider, Cline, Continue, OpenCode, or Roo Code with Anthropic, OpenAI, or Bedrock, you should check this repo. The audit findings show which specific bugs affect you: volatile-message bugs that burn cache breakpoints every turn, missing cache keys for OpenAI, opt-in defaults that leave most users with zero caching, and Bedrock custom-ARN gaps. Run the Python verification script against your agent's requests and see if your cache hit rate is zero.

Verdict

The concept is solid and the documentation is thorough, but 16 stars and no visible test suite reflect an early-stage project. The credibility score of 0.9% matches that picture. If you're currently burning API credits on one of these harnesses, the potential savings are real. Just go in knowing you're adopting something pre-1.0 with limited community validation.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.