Open Source Prompt Engineering Projects
Prompt engineering is the practice of designing and refining instructions and context so language models produce useful, reliable responses. It addresses challenges such as ambiguous requests, inconsistent output formats, missed constraints, and poor performance on specialized tasks. Techniques range from structuring a single prompt to using examples, reusable templates, and evaluation to improve results across different models and workflows.
Open source tools in this area include prompt libraries, editors, testing and evaluation frameworks, red-teaming utilities, and systems for assembling prompts within applications or agent workflows. When choosing a tool, consider its license, maintenance activity, supported models, setup requirements, and fit with your development stack. Prompt engineering resources are useful to developers, researchers, and teams building or evaluating applications powered by language models.
24 repositories · updated October 4, 2026
session-to-skill: Turn OpenCode Sessions into Reusable Skills
An OpenCode skill that turns shared session links into reusable, publishable AI agent skills. It analyzes session patterns, checks the generated skill against TRACE and security criteria, then packages the result as a ZIP.
context-engineering-kit: Improve AI Coding Agent Workflows
A collection of installable skills, commands, and agents for improving coding-agent reliability and code quality. Use it to add reflection, specification-driven development, multi-agent workflows, and code review to supported AI coding tools.
education-agent-skills: Evidence-Grounded Skills for Education AI
A library of 165 structured, evidence-grounded skills for education tasks, designed for AI agents and manual use. It supports teachers, school leaders, curriculum designers, and EdTech builders across planning, assessment, learning science, and student study support.
EasyInstruct: Generate, Select, and Prompt LLM Instructions
EasyInstruct is a Python framework for preparing instruction data and prompts for large language model research. It combines instruction generation and dataset selection tools with prompt and local-model execution modules.
promptbench: Evaluate LLMs and Test Prompt Robustness
PromptBench is a Python library for evaluating language and multimodal models across datasets, prompting methods, and adversarial attacks. It suits researchers and developers comparing model behavior or studying robustness and dynamic evaluation.
llm-guard: Add Security Checks to LLM Interactions
LLM Guard is a Python toolkit for screening prompts and model outputs for risks such as prompt injection, harmful content, and sensitive data. It is archived, so consider it for existing integrations or evaluation, not as a maintained security layer.
loop-engineering: Design and Operate AI Coding-Agent Loops
A TypeScript toolkit and pattern library for designing repeatable workflows around AI coding agents. It helps teams discover work, delegate tasks, verify results, and preserve state instead of managing every step through prompts.
spacy-llm: Add LLM Tasks to spaCy NLP Pipelines
spacy-llm connects large language models to spaCy pipelines, turning model responses into structured NLP outputs without training data. It suits teams prototyping NLP tasks or combining LLM components with conventional spaCy processing.
loopy: Find and Run Repeatable AI-Agent Workflows
Loopy is an installable skill for discovering, adapting, creating, and running bounded AI-agent workflows called loops. It works with the separate Loop Library catalog and suits people who want repeatable tasks with explicit checks and stopping points.
get-shit-done: Structure Claude Code Work with Prompts and Specs
get-shit-done was a lightweight system for meta-prompting, context engineering, and spec-driven development with Claude Code. This repository is archived; development has moved to GSD Core in the Open GSD repository.
humanizer: Edit AI-Written Text into Plainer Prose
Humanizer is an agent skill for revising AI-written text while preserving its meaning and factual details. Use it in supported coding agents or other skill-compatible tools when you want clearer, less formulaic prose.
andrej-karpathy-skills: Guide Claude Code Toward Safer Changes
A compact set of coding guidelines for Claude Code, based on Andrej Karpathy’s observations about common LLM coding mistakes. It encourages clarification, simpler implementations, focused edits, and verifiable goals.