One Source of Truth for All Your AI Coding Assistants
Write once. Compile to 37 AI coding agents: GitHub Copilot, Claude Code, Cursor, and more.
Try Playground · Documentation · Quick Start · VS Code Extension
- ❌ Even 2–3 AI tools × many repos = dozens of config files drifting out of sync
- ❌ Security policy update? Manual changes across every repo, for every tool
- ❌ Switching from Copilot to Codex? Rewrite instructions, READMEs, and docs everywhere
- ❌ No audit trail, no inheritance, no validation
- ✅ Write once in
.prs→ compile to all 37 agents - ✅ Update the source → propagates everywhere automatically
- ✅ Hierarchical inheritance like code, not copy-paste
- ✅ Full validation, audit trail, and version pinning
npm install -g @promptscript/cli
prs init # auto-detects your tech stack
prs compile # outputs to all AI toolsdocker run --rm -v $(pwd):/workspace ghcr.io/mrwogu/promptscript:latest compileOpen Playground - no install needed.
prs hooks install # auto-compiles on .prs save, protects generated filesFrom now on, every time you edit a .prs file, outputs recompile automatically. AI agents are blocked from overwriting generated configs.
After prs compile, a PromptScript language skill is automatically injected into your AI agents. They learn the .prs syntax and can create, edit, and manage your PromptScript files for you. Just ask your agent to add a new rule, change a standard, or create a shortcut — it already knows how.
Source: .promptscript/project.prs
@meta { id: "checkout-service" syntax: "1.0.0" }
@inherit @company/backend-security
@use @fragments/testing
@use @fragments/typescript-strict
@identity {
"""
You are an expert Backend Engineer working on the Checkout Service.
This service handles payments using hexagonal architecture.
"""
}
@shortcuts {
"/review": "Security-focused code review"
"/test": "Write unit tests with Vitest"
"/migrate": "Generate Prisma migration"
}
Run: prs compile generates native config files for every AI tool:
📄 .github/copilot-instructions.md
📄 CLAUDE.md
📄 .cursor/rules/project.mdc
📄 .agent/rules/project.md
📄 AGENTS.md
📄 OPENCODE.md
📄 GEMINI.md
... and 30 more agent formats
Example output - the generated CLAUDE.md:
# CLAUDE.md
You are an expert Backend Engineer working on the Checkout Service.
This service handles payments using hexagonal architecture.
## Code Style
- Use strict mode
- Prefer interfaces over types
- Never use `any` type
## Commands
- `/review` - Security-focused code review
- `/test` - Write unit tests with Vitest
- `/migrate` - Generate Prisma migrationOne .prs file. Every AI tool gets native, idiomatic output. No manual formatting.
Hierarchical Inheritance - compose standards from organization → team → project level, just like code:
@inherit @company/global-security
@inherit @team/backend-standards
@extend @standards.testing { coverage: "95%" }
Parameterized Templates - reusable stacks with typed parameters, like Infrastructure as Code:
@inherit @stacks/typescript-service(projectName: "checkout", port: 8080)
Skills - define reusable AI skills with SKILL.md files, resource bundles, and input/output contracts. Compile them to native skill formats for Claude Code, Copilot, Cursor, and more:
@skills {
deploy: {
description: "Deploy service to production"
userInvocable: true
allowedTools: ["Bash", "Read"]
}
}
Registry Resolver - import from any Git repository with Go-style URL imports or short aliases. Auto-discovers skills from repos that don't have .prs files. Lockfile support for reproducible builds. Vendor mode for offline/air-gapped CI:
# Alias (configured once in promptscript.yaml)
@use @company/security
# Or direct URL import — no config needed
@use github.com/acme/shared-standards/@fragments/security@^1.0.0
# promptscript.yaml
registries:
'@company': github.com/acme/promptscript-baseAI-Assisted Migration - already have CLAUDE.md or .cursorrules? Convert automatically:
prs init --migrateWatch Mode - auto-recompile on every change:
prs compile --watchZero Learning Curve - A PromptScript language skill is automatically compiled into your AI agents' native skill format. Your agents learn the syntax, so they manage your .prs files — you just tell them what you want in plain language.
Docker CI/CD - validate in any pipeline:
docker run --rm -v $(pwd):/workspace ghcr.io/mrwogu/promptscript:latest validate --strictPromptScript compiles to native config formats for every major AI coding agent:
- 🟢 GitHub Copilot →
.github/copilot-instructions.md - 🟢 Claude Code →
CLAUDE.md,.claude/skills/*.md - 🟢 Cursor →
.cursor/rules/*.mdc - 🟢 Google Antigravity →
.agent/rules/*.md - 🟢 Factory AI →
AGENTS.md - 🟢 OpenCode →
OPENCODE.md - 🟢 Gemini CLI →
GEMINI.md
Plus 30 more: Windsurf, Cline, Roo Code, Codex, Continue, Augment, Goose, Kilo Code, Amp, Trae, Junie, Kiro CLI, and others. Each outputs to its native config path. See the full list of formatters.
- 🔒 Private registries - host standards on internal Git repos
- 📌 Version pinning -
@inherit @company/security@2.1.0 - ✅ CI validation -
prs validate --strict --output json - 📋 Full audit trail - all changes tracked in version control
| Resource | Description |
|---|---|
| Getting Started | 5-minute quickstart guide |
| Language Reference | Full syntax documentation |
| Guides | Inheritance, registry, migration, and more |
| Enterprise | Scaling across organizations |
| VS Code Extension | Syntax highlighting for .prs files |
Contributions welcome! See CONTRIBUTING.md. | Full Roadmap →
