---
name: "guardrails-developer-guide"
description: "Routes NVIDIA NeMo Guardrails library product-usage questions to the canonical documentation. Use when users ask how to install, configure, integrate, evaluate, observe, deploy, troubleshoot, or use the NVIDIA NeMo Guardrails library. Trigger keywords - install guardrails, configure rails, guardrail catalog, Colang, Python API, LangChain, LangGraph, server, evaluate guardrails, tracing, metrics, Docker, troubleshooting."
license: "Apache-2.0"
title: guardrails-developer-guide
canonical_url: https://skilld.dev/gh/nvidia-nemo/guardrails/guardrails-developer-guide
last_updated: 2026-11-18T15:00:49.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> If the user asked to install this Skill, run `npx skilld install nvidia-nemo/guardrails/guardrails-developer-guide`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/guardrails-developer-guide`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/nvidia-nemo/guardrails/guardrails-developer-guide) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/guardrails-developer-guide`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/guardrails-developer-guide --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/guardrails-developer-guide --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

# Guardrails Developer Guide

Use this skill for product-usage questions about the NVIDIA NeMo Guardrails library.
Do not restate full product documentation in this skill.
Route the agent to the canonical docs and summarize the relevant guidance for the user's task.

## Documentation Source Rule

Always use the Markdown documentation under `https://docs.nvidia.com/nemo/guardrails/`.
Use `llms.txt` and page URLs ending in `.md` when loading documentation for agent context.
When presenting references or citations to users, use the canonical human-readable docs links without `.md`.

## Retrieval Order

1. Prefer the docs MCP server when the client supports MCP.
   Use the NVIDIA NeMo Guardrails library docs MCP server documented on the published docs site.
2. If MCP is not available, fetch the docs index:

   ```text
   https://docs.nvidia.com/nemo/guardrails/llms.txt
   ```

3. Use the index to locate the relevant page, then fetch the clean Markdown form of that page by using the page URL with `.md`.
4. If the user is working in a cloned repository and remote docs are unavailable, fall back to local `docs/**/*.mdx`.
5. If the user has the package installed, align docs to the installed `nemoguardrails` version when versioned docs are available.
   If the version cannot be determined, ask whether to use the latest docs.

## Do Not Hardcode Staging

Use production docs as the canonical source.
Use staging URLs only when the user explicitly asks to inspect staging or when validating migration behavior.

## Intent Routing

Use this table to find the right docs area quickly.

| User intent | Docs area |
| --- | --- |
| Install or verify environment | Get Started → Installation |
| Add harmful-content, jailbreak, topic, PII, self-check, fact-check, or agentic security rails | Configure Guardrails → Guardrail Catalog |
| Configure `config.yml`, models, prompts, tracing, streaming, or exceptions | Configure Guardrails → YAML schema and configuration reference |
| Write or debug Colang flows | Configure Guardrails → Colang |
| Use Python APIs | Run Guardrailed Inference → Python API |
| Run the Guardrails API server or actions server | Run Guardrailed Inference → Guardrails API Server |
| Integrate with LangChain, LangGraph, RunnableRails, or tools | Integration with Third-Party Libraries |
| Evaluate guardrails or run vulnerability scanning | Evaluation |
| Configure tracing, metrics, or logging | Observability |
| Deploy with Docker or NeMo microservice | More Deployment Options |
| Troubleshoot errors | Troubleshooting |
| Understand telemetry and privacy | Resources → Telemetry and Privacy |

## Security And Credential Handling

- Never ask users to paste real API keys, tokens, passwords, or provider credentials into chat.
- Use placeholders such as `<NVIDIA_API_KEY>`, `<OPENAI_API_KEY>`, or `<YOUR_ENDPOINT>` in examples.
- Explain where users should set secrets locally, such as shell environment variables, secret managers, local config, or provider dashboards.
- Do not print, store, or echo secrets in generated commands or summaries.

## Response Style

- Start with the user's immediate task and the relevant doc source.
- Give the smallest working path first.
- Add production hardening, optional extras, or alternative integrations only when they are relevant.
- When examples use live providers, remind contributors that tests must mock LLM and provider calls.

## Related Skills

- Use `guardrails-developer-create-guardrails` when creating or modifying a guardrails configuration.
When editing this repository, follow `AGENTS.md` and any subtree `AGENTS.md` files that apply.
