awesome ai system prompts

Report Abuse

Basic Information

This repository is a curated reference and guide for crafting effective system prompts for agentic AI systems. It collects real-world system prompts, schemas, and patterns from a variety of agents and platforms including Vercel v0, same.new, Manus, OpenAI ChatGPT, Claude, Cline, Bolt.new and others. The README synthesizes principles such as clear role definition, structured instructions, explicit tool integration, iterative agent loops, environment grounding, domain constraints, safety and refusal protocols, and consistent tone. The content is organized into case studies, modular prompt examples and annotated best practices so builders and prompt engineers can study how production agents specify tools, plan actions, format outputs, and enforce safety. The repo is aimed at people designing or improving agent behaviors rather than at end users seeking a single bot.

Links

App Details

Features
A prominent feature is annotated case studies that illustrate prompt design patterns for different agent archetypes such as UI generation, pair programming, and general-purpose sandbox agents. The collection includes concrete tool schemas and examples (JSON, TypeScript-like definitions, MDX/XML tag formats), explicit agent loop definitions, thinking and planning phases, environment and system information snippets, and domain-specific style and coding constraints. It documents safety and refusal protocols and examples of tone and persona settings. Prompts are shown in both monolithic and modular layouts with references to external files like tools.json and functions-schema.json. The README synthesizes recurring best practices and contrasts architectural conventions across multiple implementations.
Use Cases
The repository helps prompt engineers, AI researchers and agent builders learn proven conventions and reuseable patterns when designing system prompts and tool integrations. It provides ready-made examples and templates for defining roles, tool calling formats, function schemas, and refusal policies that can be adapted to a new agent. Case studies demonstrate how to ground agents in environment details, enforce iterative planning and waiting for results, and apply domain-specific constraints and style guides. The material aids auditing and comparison of prompt strategies, speeds prototyping by offering explicit output formats and tags, and informs safer, more predictable agent behavior by highlighting alignment and safety protocols used in deployed systems.

Please fill the required fields*