AI Safety Guardrails

Input/output safety filters for AI systems. Pluggable filter pipeline that sits between user and LLM boundaries, inspecting and filtering both directions. Ships with prompt-injection detection, PII redaction, content-policy enforcement, and token-limit guards. The filter registry pattern makes it trivial to add custom safety filters.

Name	`guardrails`
Version	`0.1.0`
Category	ai-systems
License	Apache-2.0
Persona	engineering
Tags	ai, safety, guardrails, filters, typescript
Source	`templates/guardrails`

Render it

Three front doors, one catalog. Pick whichever suits the caller.

# CLI
npx @nanohype/sdk render guardrails --out ./my-app

# SDK
import { LocalSource, renderTemplate } from "@nanohype/sdk";
const result = await renderTemplate(source, "guardrails", variables);

# MCP — from an agent
get_template({ name: "guardrails" })

Prerequisites

Tool	Version	Why
`node`	`>=22`	Node.js runtime for the guardrails module

Variables

Required

ProjectName string: Kebab-case project name, used as package name and directory Must be lowercase kebab-case starting with a letter.

Optional

Description string — defaults to AI safety guardrails: Short project description for package.json and README
IncludePii bool — defaults to true: Include PII detection filter
IncludeContentPolicy bool — defaults to true: Include content policy filter
IncludeTests bool — defaults to true: Include test suite

What it produces

22 files. Placeholder names such as __APP_NAME__ are what the renderer substitutes into.

.env.example
.gitignore
biome.json
package.json
README.md
src/guardrails/bootstrap.ts
src/guardrails/filters/__tests__/pii.test.ts
src/guardrails/filters/__tests__/pipeline.test.ts
src/guardrails/filters/__tests__/prompt-injection.test.ts
src/guardrails/filters/content-policy.ts
src/guardrails/filters/index.ts
src/guardrails/filters/pii.ts
src/guardrails/filters/prompt-injection.ts
src/guardrails/filters/registry.ts
src/guardrails/filters/token-limit.ts
src/guardrails/filters/types.ts
src/guardrails/index.ts
src/guardrails/pipeline.ts
src/guardrails/types.ts
src/logger.ts
tsconfig.json
vitest.config.ts

Composes with

Pairs with agentic-loop
Pairs with mcp-server-ts
Pairs with ts-service
Pairs with rag-pipeline
Nests inside monorepo

Composites that use it

Enterprise AI Infrastructure — Full enterprise AI stack with agents, tool servers, safety guardrails, evaluation, prompt management, observability, and Kubernetes deployment..
AI Evaluation Suite — Standalone evaluation infrastructure with test harness, versioned prompts, and guardrails validation.
Multi-Agent System — Multi-agent architecture with A2A protocol peers, MCP tool servers, and an orchestrating agent.
Safe AI Agent — AI agent with safety guardrails, evaluation harness, and prompt management.