PromptScript¶

One source. Every AI agent.

Prompt-as-code with inheritance, validation, and a versioned registry. Compiled to 37 targets.

Claude Code Cline Codex Continue GitHub Copilot Roo Code Windsurf Cursor Amp Augment Antigravity Goose Junie Factory AI Kilo Code Kiro CLI Gemini CLI Trae Adal OpenCode CodeBuddy Command Code Claude Code Cortex Crush GitHub Copilot iFlow Kode Cursor MCPJam Mistral Vibe Antigravity Mux Neovate Factory AI OpenClaw OpenHands Gemini CLI Pi Pochi OpenCode Qoder Qwen Code Zencoder Claude Code Cline Codex Continue GitHub Copilot Roo Code Windsurf Cursor Amp Augment Antigravity Goose Junie Factory AI Kilo Code Kiro CLI Gemini CLI Trae Adal OpenCode CodeBuddy Command Code Claude Code Cortex Crush GitHub Copilot iFlow Kode Cursor MCPJam Mistral Vibe Antigravity Mux Neovate Factory AI OpenClaw OpenHands Gemini CLI Pi Pochi OpenCode Qoder Qwen Code Zencoder

Get Started in 5 Minutes See All 37 Formatters Enterprise Guide GitHub

npm install -g @promptscript/cli

VS Code Extension available →

Watch Introduction

9 lines → 37 agents

prs compile

Awaiting compilation...

Claude 4 files Copilot 4 files Codex 1 file OpenCode 3 files Gemini 3 files Factory 2 files Cursor 2 files

instructions.md review.prompt migrate.prompt security.skill

# Checkout Service v2.1

← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Tech Stack
TypeScript 5.x · Node 20+ · PostgreSQL
Prisma ORM · Zod · Vitest

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation
✓ 90% test coverage target
✓ Conventional Commits

## Security
⛔ Never log PII or card data
⛔ Parameterized SQL only
⛔ No secrets in source code

## Commands
/review Security code review
/migrate Safe DB migration

## Skills
► security-audit OWASP scan

---
mode: agent
tools: [codebase, terminal]
description: Security code review
---

Perform security-focused code review
on the Checkout Service.

Check for:
• SQL injection vulnerabilities
• XSS attack vectors
• Exposed secrets and API keys
• Input validation gaps
• IDOR vulnerabilities

---
mode: agent
tools: [terminal, codebase]
description: Database migration
---

Generate Prisma migration safely.

Steps:
1. Analyze schema changes
2. Generate migration file
3. Review for data loss risks
4. Create rollback plan
5. Validate with dry run

---
name: security-audit
description: OWASP vulnerability scan
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure
• Security misconfiguration
• Insecure dependencies

CLAUDE.md /review /migrate security.skill

# Checkout Service v2.1

← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Tech
TypeScript 5.x · Node 20+ · Prisma
Zod validation · Vitest

## Commands
/review Security code review
/migrate Safe DB migration

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation
✓ Conventional Commits

## Don'ts
✗ Don't log PII or card data
✗ Don't use raw SQL queries
✗ Don't commit secrets

Perform security-focused code review
on the Checkout Service.

Check for:
• SQL injection vulnerabilities
• XSS attack vectors
• Exposed secrets and API keys
• Input validation gaps
• IDOR vulnerabilities

Generate Prisma migration safely.

Steps:
1. Analyze schema changes
2. Generate migration file
3. Review for data loss risks
4. Create rollback plan
5. Validate with dry run

---
name: security-audit
description: OWASP vulnerability scan
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure
• Security misconfiguration

checkout.mdc security.mdc

---

description: Checkout Service v2.1
globs: ["src/**/*.ts"]
alwaysApply: true
---

# Checkout Service
← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Stack
TypeScript 5.x · Node 20+ · Prisma
Zod validation · Vitest

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation
✓ 90% test coverage

## Security
⛔ Never log PII or card data
⛔ Parameterized queries only
⛔ No secrets in source code

---
description: OWASP vulnerability scan
globs: ["src/**/*.ts"]
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure
• Security misconfiguration
• Insecure dependencies

AGENTS.md

# Checkout Service v2.1

← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Tech
TypeScript 5.x · Node 20+ · Prisma
Zod validation · Vitest

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation
✓ Conventional Commits

## Security
⛔ Never log PII or card data
⛔ Parameterized queries only
⛔ No secrets in source code

AGENTS.md security.skill

# Checkout Service Agent v2.1

← @acme/platform
← @stacks/node-typescript

## Role
Senior Engineer specializing in the
Checkout Service.

## Commands
/review Security code review
/migrate Safe DB migration

## Handoffs
➜ security-audit OWASP scan

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation

---
name: security-audit
description: OWASP vulnerability scan
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure
• Security misconfiguration

OPENCODE.md /review security.skill

# Checkout Service v2.1

← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Tech
TypeScript 5.x · Node 20+ · Prisma

## Commands
/review Security code review
/migrate Safe DB migration

## Skills
▶ security-audit OWASP scan

## Standards
✓ Strict mode, no `any`
✓ Zod for all input validation

Perform security-focused code review
on the Checkout Service.

Check for:
• SQL injection vulnerabilities
• XSS attack vectors
• Exposed secrets and API keys
• Input validation gaps
• IDOR vulnerabilities

---
name: security-audit
description: OWASP vulnerability scan
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure

GEMINI.md review.toml security.skill

# Checkout Service v2.1

← @acme/platform
← @stacks/node-typescript

## Identity
Senior Engineer specializing in the
Checkout Service.

## Tech
TypeScript 5.x · Node 20+ · Prisma

## Commands
/review Security code review
/migrate Safe DB migration

## Don'ts
✗ Don't log PII or card data
✗ Don't use raw SQL queries
✗ Don't commit secrets

[command]
name = "review"
description = "Security code review"

[command.steps]
action = "review"

# Check for:
# SQL injection vulnerabilities
# XSS attack vectors
# Exposed secrets/API keys
# Input validation gaps

---
name: security-audit
description: OWASP vulnerability scan
---

# Security Audit

Run a comprehensive OWASP Top 10 scan.

Scan for:
• Injection (SQL, NoSQL, OS)
• Broken authentication
• Sensitive data exposure

Why PromptScript?

Enterprise-grade features for managing AI instructions at scale

Write Once, Compile to 37 AI Agents

Define standards in PromptScript. Compile to GitHub Copilot, Claude Code, Cursor, Windsurf, Cline, Roo Code, and 31 more. Add new agents without rewriting.

37 agents and counting - see all supported formatters

Hierarchical Inheritance at Scale

Org-wide policies inherit into team standards, which inherit into projects. Change once at the top, propagate everywhere.

Audit Trails & Policy Enforcement

Every standard versioned and traceable. Run prs validate --strict in CI to block non-compliant configs before production.

Your Private Standards Registry

Host standards in a private Git registry. Teams inherit from @company/security, @company/react-app - your standards, your control.

CI/CD Pipeline Integration

Add prs compile --check to your pipeline. Detect drift, enforce standards, fail builds on non-compliance.

Zero-Touch Rollout to 100+ Repos

Update a security policy once. Every repo gets the latest standards compiled to every AI tool - no manual PRs.

Zero Learning Curve

A PromptScript language skill is automatically compiled into your AI agents. They learn the syntax and manage your .prs files for you — just ask in plain language.

Import from Anywhere

Use packages from Git repositories - or import skills directly from any AI coding plugin

# Company-wide security rules
@use @company/security

# Team-specific standards
@use @company/standards/code-quality

# Open-source skills (auto-discovered from SKILL.md)
@use github.com/some-org/claude-skills/skills/tdd-workflow

# Pin to a version
@use @company/security@^1.0.0

Configure aliases once, use everywhere:

# promptscript.yaml
registries:
  '@company': github.com/acme/promptscript-base
  '@team': github.com/acme/team-frontend

Zero infrastructure required - uses Git directly. Private repo? Your existing SSH keys and tokens just work.

Registry Guide See Examples

The Problem: AI Instructions Don't Scale

Your organization adopted AI coding tools. Without infrastructure, every team configures differently.

Scale Problem

Updating a security policy across 100 microservices takes weeks of manual PRs. One change = 100 commits.

Tool Fragmentation

Your teams use Copilot, Claude, and Cursor. Each has its own format. Keeping them in sync is a full-time job.

Governance Void

No audit trail. Junior devs miss critical security context. Unvetted instructions everywhere.

Vendor Lock-in Risk

You went all-in on Cursor rules. Now leadership evaluates Claude Code. Rewriting 200 repos is not an option.

⚠️ Result: Inconsistent code quality, security risks, and operational chaos.

The Solution: PromptOps

PromptOps is the practice of managing AI instructions as versioned, auditable infrastructure. Treat your AI instructions as managed infrastructure - compiled, validated, and deployed.

flowchart TB
    subgraph org["🏢 Organization Registry"]
        direction TB
        OrgSec["@company/security<br/>(Security team)"]
        OrgTS["@company/typescript<br/>(Platform team)"]
        OrgJava["@company/java<br/>(Platform team)"]
    end

    subgraph teams["👥 Team Standards"]
        direction TB
        Backend["@backend/standards"]
        Frontend["@frontend/standards"]
        Data["@data/standards"]
    end

    subgraph projects["📦 Project Repos (100+)"]
        direction TB
        P1["checkout-service/.prs"]
        P2["user-api/.prs"]
        P3["dashboard/.prs"]
        Pn["...more repos"]
    end

    OrgSec --> Backend & Frontend & Data
    OrgTS --> Backend & Frontend
    OrgJava --> Data

    Backend --> P1 & P2
    Frontend --> P3
    Data --> Pn

    subgraph compile["⚙️ prs compile"]
        Compiler
    end

    P1 & P2 & P3 --> Compiler

    Compiler --> Copilot["GitHub Copilot"]
    Compiler --> Claude["Claude Code"]
    Compiler --> Cursor["Cursor"]
    Compiler --> Antigravity["Antigravity"]
    Compiler --> Factory["Factory AI"]
    Compiler --> OpenCode["OpenCode"]
    Compiler --> Gemini["Gemini CLI"]
    Compiler --> CI["CI/CD Audit"]

See It In Action

Write once, compile to all major AI tools

PromptScript (.prs)

@meta { id: "checkout-service" syntax: "1.0.0" }

@inherit @company/backend-standards

@identity {
  """
  You are an expert Fullstack Engineer working on the Checkout Service.
  This service handles payments using hexagonal architecture.
  """
}

@standards {
  typescript: [
    "Strict mode enabled",
    "Never use any type"
  ]
  testing: [
    "Use vitest as test framework",
    "Maintain 90% code coverage"
  ]
}

@restrictions {
  - "Never expose API keys in code"
  - "Never skip input validation"
}

@shortcuts {
  "/review": "Security-focused code review"
  "/test": "Write unit tests with Vitest"
}

@skills {
  security-audit: {
    description: "Audit code for vulnerabilities"
    content: "Check for: SQL injection, XSS, IDOR, PII leaks."
  }
}

GitHub CopilotClaude CodeCursorAntigravity

Generated file tree:

.github/
├── copilot-instructions.md
├── prompts/
│   ├── review.prompt.md
│   └── test.prompt.md
└── copilot-skills/
    └── security-audit.md

.github/copilot-instructions.md

# GitHub Copilot Instructions

> Auto-generated from project.prs

## Project
You are an expert Fullstack Engineer working on the Checkout Service.
This service handles payments using hexagonal architecture.

## Code Standards
- TypeScript: strict mode, no `any` types
- Testing: vitest, 90% coverage

## Restrictions
- Never expose API keys in code
- Never skip input validation

<!-- Inherited from @company/backend-standards -->

.github/prompts/review.prompt.md

---
mode: agent
description: Security-focused code review
---

Perform a security-focused code review on the selected code.
Check for: SQL injection, XSS, IDOR, PII leaks.

.github/prompts/test.prompt.md

---
mode: agent
description: Write unit tests with Vitest
---

Write comprehensive unit tests for the selected code using Vitest.
Target coverage: 90%.

.github/copilot-skills/security-audit.md

---
name: security-audit
description: Audit code for vulnerabilities
---

# Security Audit Skill

Check for: SQL injection, XSS, IDOR, PII leaks.

Generated file tree:

CLAUDE.md
.claude/
└── skills/
    ├── review.md
    ├── test.md
    └── security-audit.md

CLAUDE.md

# CLAUDE.md

## Project
You are an expert Fullstack Engineer working on the Checkout Service.
This service handles payments using hexagonal architecture.

## Code Style
- Strict TypeScript, no `any`
- Testing: vitest, >90% coverage

## Don'ts
- Don't expose API keys in code
- Don't skip input validation

## Commands
/review - Security-focused code review
/test - Write unit tests with Vitest

.claude/skills/review.md

---
name: review
description: Security-focused code review
---

Perform a security-focused code review.
Check for: SQL injection, XSS, IDOR, PII leaks.

.claude/skills/test.md

---
name: test
description: Write unit tests with Vitest
---

Write comprehensive unit tests using Vitest.
Target coverage: 90%.

.claude/skills/security-audit.md

---
name: security-audit
description: Audit code for vulnerabilities
---

Check for: SQL injection, XSS, IDOR, PII leaks.

Generated file tree:

.cursor/
└── rules/
    ├── checkout-service.mdc
    └── security-audit.mdc

.cursor/rules/checkout-service.mdc

---
description: Checkout Service standards
globs: ["src/**/*.ts"]
alwaysApply: true
---

You are working on the Checkout Service.
This service handles payments using hexagonal architecture.

## Code Style
- TypeScript: strict mode, no `any`
- Testing: vitest, 90% coverage

## Never
- Expose API keys in code
- Skip input validation

.cursor/rules/security-audit.mdc

---
description: Security audit skill
globs: ["src/**/*.ts"]
---

# Security Audit

Check for: SQL injection, XSS, IDOR, PII leaks.

Generated file tree:

.agent/
└── rules/
    ├── checkout-service.md
    └── security-audit.md

.agent/rules/checkout-service.md

---
description: Checkout Service standards
activation: always
---

You are working on the Checkout Service.
This service handles payments using hexagonal architecture.

## Code Style
- TypeScript: strict mode, no `any`
- Testing: vitest, 90% coverage

## Never
- Expose API keys in code
- Skip input validation

.agent/rules/security-audit.md

---
description: Security audit skill
activation: manual
---

# Security Audit

Check for: SQL injection, XSS, IDOR, PII leaks.

Before & After

See the difference PromptScript makes

Without PromptScriptWith PromptScript

Maintain separate config files for each AI tool in every repo:

.github/copilot-instructions.md   # GitHub Copilot
CLAUDE.md                          # Claude Code
.cursor/rules/project.mdc         # Cursor
AGENTS.md                          # Codex / Factory
GEMINI.md                          # Gemini CLI

No validation - instructions drift silently
No inheritance - copy-paste across repos
No audit trail - who changed what?
Switching tools = rewrite instructions, READMEs, and docs everywhere

Write one .prs file, compile to all 37 agents:

@meta { id: "my-project" syntax: "1.0.0" }
@inherit @company/standards
@identity { """You are an expert developer.""" }

prs compile  # Generates all 37 output files

Validated at compile time and in CI/CD
Hierarchical inheritance from org → team → project
Full audit trail via Git
New tools supported automatically

See It Live

Try PromptScript right in your browser - no installation needed

Open the Playground

Video walkthrough coming soon

Join the Community

PromptScript is open source and early-adopter friendly

We're building the future of AI instruction management in the open.

Star us on GitHub Report Issues

Ready to Standardize Your AI Instructions?

Join engineering teams using PromptScript to manage AI context at scale

Get Started in 5 Minutes Enterprise? Let's Talk View on GitHub