nemo-guardrails

NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.

Quality

61%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/safety-alignment/nemo-guardrails/SKILL.md

Quality

Content

57%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is action-oriented with concrete, mostly-executable NeMo Guardrails examples and useful troubleshooting, but it is volume-heavy with five inlined workflows, several examples depend on undefined helper functions, and the referenced deeper-detail files are missing from the bundle.

Suggestions

Create the referenced files (references/colang-guide.md, integrations.md, performance.md) or remove the dangling links so progressive disclosure navigation works.

Define or stub the helper functions used in workflows 2–3 (toxicity_detector, extract_facts, verify_facts, fact_check_action) so the code is genuinely copy-paste runnable.

Move two or three of the five full workflow examples into a reference file and keep only a representative quick-start plus pointers in SKILL.md to reduce token weight.

Dimension	Reasoning	Score
Conciseness	The body is mostly practical code rather than concept explanation, but five full inline workflow code blocks plus alternatives/issues/hardware/resources sections make it substantially heavy, and much of the workflow detail could be trimmed or moved into the referenced files.	3 / 5
Actionability	Concrete, mostly executable code shows real API usage (RailsConfig.from_content, LLMRails, rails.generate) with expected outputs, but workflows 2–3 rely on undefined helpers (toxicity_detector, extract_facts, verify_facts, fact_check_action), preventing full copy-paste readiness.	4 / 5
Workflow Clarity	The five workflows are clear individual recipes (define config → create rails → generate), but they are presented as parallel use cases rather than a coherent sequenced process, and explicit validation checkpoints appear in only two of the five flows.	3 / 5
Progressive Disclosure	References are clearly signaled and one level deep in the Advanced topics section, but the referenced files (references/colang-guide.md, integrations.md, performance.md) do not exist in the bundle and the five full workflows are inlined rather than split out, so navigation does not fully deliver.	3 / 5
	Total	13 / 20 Passed

Description

66%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description clearly communicates what the skill does via a specific, well-targeted feature list, but it omits any explicit 'Use when...' trigger guidance, which caps completeness. Trigger-term coverage is strong but missing common synonyms like 'guardrails' and 'content moderation'.

Suggestions

Add an explicit 'Use when...' clause, e.g. 'Use when adding runtime safety rails to LLM applications, or when the user mentions guardrails, jailbreak detection, PII filtering, or content moderation.'

Include the natural term 'guardrails' and synonyms like 'content moderation' and 'prompt injection' in the description body so users' phrasing matches the trigger.

Drop the marketing-style 'Production-ready, runs on T4 GPU' over-claim from the description; move hardware specifics to the body.

Dimension	Reasoning	Score
Specificity	Lists six concrete named mechanisms ("jailbreak detection", "input/output validation", "fact-checking", "hallucination detection", "PII filtering", "toxicity detection") and the Colang 2.0 DSL, but it reads as a feature list rather than verbatim actions and "Production-ready, runs on T4 GPU" is an over-claim, keeping it below a 5.	4 / 5
Completeness	The "what" is clearly stated ("NVIDIA's runtime safety framework for LLM applications. Features ...") but there is no "Use when..." clause or equivalent explicit trigger guidance, so per the guideline completeness is capped at 3.	3 / 5
Trigger Term Quality	Good natural keywords users would say ("jailbreak detection", "PII filtering", "toxicity detection", "fact-checking", "hallucination detection"), but common synonyms like "guardrails", "content moderation", and "prompt injection" are absent, so it is not comprehensive.	4 / 5
Distinctiveness Conflict Risk	The named NVIDIA/NeMo Guardrails niche with Colang 2.0 and six specific mechanisms is mostly distinct, with only minor overlap risk against a generic LLM-safety or content-moderation skill.	4 / 5
	Total	15 / 20 Passed

Validation

81%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 3 missing	Warning
referenced_paths_exist	Referenced path issues: 6 missing	Warning

	Total	13 / 16 Passed

Repository: OpenLAIR/dr-claw
Path: skills/safety-alignment/nemo-guardrails/SKILL.md
Commit: 4b35756

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.