AI analysis grounded in the code graph — computed facts, not vibes · 2026-07-06T05:40:57Z
This repository is a curated, crowdsourced collection of extracted "system prompts" — the hidden instructions that configure large language model chatbots and coding assistants — organised by vendor (Anthropic, OpenAI, Google, xAI, Microsoft, Perplexity, and others). Content is almost entirely Markdown documentation files per model/product (e.g. Anthropic/claude-opus-4.8.md, OpenAI/Codex/gpt-5.5.md), with a small amount of JavaScript/Python tooling bundled inside one specific extracted skill folder. It targets developers, prompt engineers, and researchers who want to inspect or reuse how vendors configure their assistants' behaviour, tool access, and guardrails.
The GitHub metadata shows 0 stars gained this period despite a large existing base of 50,307 — growth has evidently stalled, and the evidence provided doesn't explain any recent surge. The commit history shows steady, frequent updates (new model prompts added within days of release: Claude Sonnet 5, Claude Design/Opus 4.8, GPT-5.5 Codex, Copilot for macOS), which plausibly explains historical interest, and the README cites a Washington Post feature as external validation. However, with zero star growth in this window, the cadence of updates alone isn't currently translating into new attention.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.