How your-voldemort/system-prompts-and-models-of-ai-tools works
This repository is not an executable application; it is a curated corpus of AI agent prompts, tool manifests, and related model configuration text. The architectural “system” is a content library organized by vendor/product, with Markdown and plain-text artifacts as the primary storage format.
The root README is the public entry point and marketing/documentation layer. It frames the repo as “system prompts and models of AI tools,” links to external communities and sponsors, and points readers to the collection rather than to a running service. The README also contains badges and outbound links, which are presentation-only and do not affect the repo’s internal data model.
The main structural boundary is the top-level vendor/application folders. Each folder isolates a specific assistant or product family, such as Cursor, Windsurf, Replit, Claude Code, Gemini, Lovable, Qoder, and others. This separation lets the repository compare prompt styles across products without mixing their formats or tool contracts.
Within each vendor folder, the files typically split into three kinds of artifacts: system prompts, task/agent prompts, and tool definitions. Prompts are stored as .txt files, while tool schemas are usually .json. This suggests a clear boundary between natural-language instruction layers and structured tool interfaces.
A second organizing axis is mode or capability. Many products have distinct prompt variants for chat, planning, agent mode, builder mode, spec mode, or task-specific workflows. Examples include Cursor’s multiple agent prompt versions, Kiro’s mode/classifier/spec/vibe prompts, and Traycer AI’s phase/plan mode assets. Architecturally, this means the repo captures not just products, but internal operating modes of those products.
Some folders contain multiple versions of the same prompt or tool set, indicating historical snapshots and model-specific variants. For example, there are separate prompts for different model families or release dates. This makes the repository a versioned reference archive rather than a single canonical configuration.
The “Open Source prompts” section acts as a cross-product comparison set for openly available agent systems such as Bolt, Cline, Codex CLI, Gemini CLI, Lumo, and RooCode. This is the closest thing to a normalized benchmark area in the repo, where prompts are grouped by open-source tooling ecosystem rather than by vendor leak/source.
The data flow in the repository is simple: external prompt/tool definitions are collected, categorized, and stored as text files, then exposed through the README and folder hierarchy for browsing. There is no evidence of a processing pipeline, database, API backend, or runtime service in this repo itself.
assets/ holds static presentation resources, such as the logo image referenced by the README. These assets support the documentation layer and do not participate in prompt semantics or tool definitions.
There are no application dependencies, build steps, or deployment targets visible in the tree. The “infrastructure” is therefore Git-based source control and static file hosting, with external links to Discord, Patreon, Ko-fi, and third-party badges.
Main entry points: README.md README · Cursor Prompts Cursor · Anthropic Anthropic · Google Google
Components 22
Documentation
Public entry point and repository framing
- README · Front door to the corpus and its framing
README.md
Vendor archives
Proprietary product families and their prompt/tool artifacts
- Anthropic · Claude-family prompts and tool schemas
Anthropic - Cursor · Multiple agent and chat prompt variants
Cursor Prompts - Google · Antigravity and Gemini prompt variants
Google - Kiro · Mode, spec, and vibe prompts
Kiro - Trae · Builder and chat prompt pair
Trae - Traycer AI · Phase and plan mode artifacts
Traycer AI - Amp · Model-specific prompt manifests
Amp - Replit · Prompt plus structured tool manifest
Replit - Windsurf · Wave-specific prompt and tools
Windsurf - VSCode Agent · Model-targeted prompt variants and titles
VSCode Agent - Xcode · Platform action prompts and system prompt
Xcode - Lovable · Agent prompt and tool schema
Lovable - Claude Code · Anthropic coding agent prompt and tools
Anthropic/Claude Code - Agent Prompt · Cursor agent prompt versions and CLI prompt
Cursor Prompts/Agent Prompt v1.0.txt - Agent Tools · Cursor agent tool definition set
Cursor Prompts/Agent Tools v1.0.json - Antigravity · Fast and planning modes for Google Antigravity
Google/Antigravity - Phase modes · Traycer phase-mode prompts and tools
Traycer AI/phase_mode_prompts.txt - Tools · Replit structured tool manifest
Replit/Tools.json
Open-source set
Cross-product benchmark of openly available agent prompts
- Open Source · Openly available agent prompts grouped by ecosystem
Open Source prompts - CLI prompts · Cline, Codex CLI, and Gemini CLI references
Open Source prompts/Codex CLI
Static assets
Presentation resources referenced by the README
- Assets · Presentation resources for the repository
assets
Connections 21
- README → Anthropic: indexes · README surfaces the vendor archive
- README → Cursor: indexes · README surfaces the vendor archive
- README → Google: indexes · README surfaces the vendor archive
- README → Kiro: indexes · README surfaces the vendor archive
- README → Trae: indexes · README surfaces the vendor archive
- README → Traycer AI: indexes · README surfaces the vendor archive
- README → Amp: indexes · README surfaces the vendor archive
- README → Replit: indexes · README surfaces the vendor archive
- README → Windsurf: indexes · README surfaces the vendor archive
- README → VSCode Agent: indexes · README surfaces the vendor archive
- README → Xcode: indexes · README surfaces the vendor archive
- README → Lovable: indexes · README surfaces the vendor archive
- README → Open Source: indexes · README surfaces the open-source benchmark set
- README → Assets: references · README references static assets
- Anthropic → Claude Code: contains · Claude Code folder holds prompt and tools artifacts
- Cursor → Agent Prompt: contains · Cursor prompt variants live here
- Cursor → Agent Tools: contains · Cursor tool schema complements the prompts
- Google → Antigravity: contains · Antigravity prompts are organized as modes
- Traycer AI → Phase modes: contains · Phase-mode prompt bundle
- Replit → Tools: contains · Replit couples prompts with a tool manifest
- Open Source → CLI prompts: contains · Open-source CLI family examples