"Create and validate NVIDIA NeMo Guardrails configs: config.yml,
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill configure-rails --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Configure Rails?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-configure-rails)More formats (shields.io, HTML) on the badges page.
---
name: configure-rails
description: "Create and validate NVIDIA NeMo Guardrails configs: config.yml,
Colang flows, prompts, guardrail catalog rails, custom actions, providers,
caching, knowledge base, and config migration."
disable-model-invocation: true
metadata:
disco-role: operating
license: NOASSERTION
---
# configure-rails
Use this sub-skill when the task is about authoring or validating a guardrails configuration, not running chat, server, evaluation, or source-checkout work.
## Route here
- Create or repair `config.yml`, `prompts.yml`, `rails.co`/`main.co`, `config.py`, `actions.py`, `kb/`, `import_paths`, caching, or knowledge-base settings.
- Choose Colang 1.0 vs 2.x and load configs with `RailsConfig.from_path()` or `RailsConfig.from_content()`.
- Add built-in rails from the guardrail catalog or map a third-party rail to the right `models`/`rails` entries.
- Register custom actions, action parameters, providers, embedding providers, or embedding search providers.
- Migrate older Colang/config layouts with `nemoguardrails convert`.
## Route away
- Chat/server/runtime usage, streaming, RunnableRails, or `/v1/*` calls: `../run-rails/SKILL.md`.
- Evaluation, tracing, observability, or telemetry: `../evaluate-and-observe/SKILL.md`.
- Source checkout changes, PR policy, maintainer validation, or docs generation: `../repo-development/SKILL.md`.
- Install/import discovery only: `../setup-and-basics/SKILL.md`.
## Operating workflow
1. Start with `references/configuration-and-colang.md` to choose the right folder layout, loader, and Colang version.
2. Use `references/guardrail-catalog.md` to map the desired rail family to the required flow names, prompt tasks, model types, and optional dependencies.
3. Use `references/custom-actions-and-providers.md` for `actions.py`, `config.py`, provider registration, and embedding provider registration.
4. Use `references/troubleshooting.md` whenever config loading fails or a validation rule is expected to fire.
5. Validate with `scripts/validate_config.py`; pass `--instantiate` only when you want to confirm that `LLMRails` can be constructed without generating a response.
## Safe helper
```bash
python sub-skills/configure-rails/scripts/validate_config.py --config path/to/config
python sub-skills/configure-rails/scripts/validate_config.py --config path/to/config --instantiate
```
The helper loads the config, prints a concise summary, and never calls generation, chat, or server methods.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!