[GitHub Trending] asgeirtj/system_prompts_leaks

8.2 relevance

Leaked system prompts from major AI companies; highly novel and strategically important.

AI/ML github.com

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For a platform engineer building agentic systems, these leaked prompts expose the actual behavioral constraints, tool definitions, and safety guidelines that frontier labs embed—insights you can directly apply when designing multi-agent orchestration (LangGraph, CrewAI) and crafting robust system prompts for your own AI services.

Key Takeaways

Study the Claude Code and ChatGPT Sol prompts in this repo to understand how top labs structure tool-use, subagent dispatch, and safety rules, then adapt those patterns into your agent orchestration pipelines.

Extracted system prompts from Anthropic - Claude Fable 5.1, Opus 5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-6-Astra, Codex. Google - Gemini 3.8 Flash, 3.1 Pro, Antigravity. xAI - Grok, Gro...

Summary

The asgeirtj/system_prompts_leaks repository on GitHub archives verbatim system prompts from major LLMs including ChatGPT, Claude, Gemini, and Grok, with recent additions covering Claude Fable 5.1, Gemini 3.8 Flash, and Codex GPT-6-Astra. It captures hidden instructions for coding assistants (Claude Code, Muse Code, OpenCode), integration-specific prompts (Claude Design, Cowork, Science), and injected reminders, offering a raw look at how AI behavior is shaped pre-conversation. The repo has been used by The Washington Post and CEPS' AI World for journalism and analytics, reflecting growing interest in reverse-engineering AI guardrails.

Author

asgeirtj