llamaguard

Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.

Quality

66%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./07-safety-alignment/llamaguard/SKILL.md

Quality

Content

65%

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

The content is highly actionable with executable, copy-paste-ready code, but it is verbose for the main file, lacks validation checkpoints in its batch/output workflows, and points to reference files that do not exist.

Suggestions

Create the referenced files (references/custom-categories.md, benchmarks.md, deployment.md) so the progressive-disclosure links resolve, or remove the broken links.

Add validation/verification checkpoints to the batch moderation (Workflow 3) and output-filtering workflows (e.g. confirm parsed category, retry on parse failure).

Move the full FastAPI server and NeMo Guardrails configuration examples into reference files to tighten the main SKILL.md.

Dimension	Reasoning	Score
Conciseness	The body avoids explaining concepts Claude already knows and the code is lean, but five near-complete production examples (a full FastAPI server, full vLLM batch loop, full NeMo config) make it long and padded where trimming or offloading to references would tighten it.	2 / 3
Actionability	Code throughout is fully executable and copy-paste ready — real model IDs, transformers/vLLM/FastAPI/NeMo snippets, and concrete commands — matching the 'fully executable, specific examples' anchor.	3 / 3
Workflow Clarity	Workflows are clearly labeled and sequenced, but batch operations (Workflow 3) and output-filtering lack validation/verification checkpoints; the rubric caps workflow clarity at 2 when such checkpoints are missing.	2 / 3
Progressive Disclosure	Sections are organized and reference three reference files, but those referenced files (references/custom-categories.md, benchmarks.md, deployment.md) do not exist, and substantial inline content (full API server, NeMo config) that should be split out remains in SKILL.md.	2 / 3
	Total	9 / 12 Passed

Description

67%

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 is specific and distinctive, naming concrete moderation actions and a clear niche, but lacks an explicit 'Use when...' trigger clause and some natural user phrasings. Completeness and trigger-term quality are capped accordingly.

Suggestions

Add an explicit trigger clause such as 'Use when moderating LLM inputs/outputs, filtering unsafe content, or classifying text for safety categories.'

Include natural phrasings users would actually say (e.g. 'is this content safe', 'moderate this message', 'filter unsafe prompts') to improve trigger-term coverage.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions and artifacts — "input/output filtering", the six named safety categories, and deployment targets (vLLM, HuggingFace, Sagemaker) — matching the 'lists multiple specific concrete actions' anchor.	3 / 3
Completeness	It clearly states what the skill does but has no explicit 'Use when...' trigger clause answering when Claude should invoke it; per the rubric guidelines, a missing explicit trigger caps completeness at 2.	2 / 3
Trigger Term Quality	Contains relevant keywords ("moderation model", "content moderation", "input/output filtering") but misses common natural phrasings users would say (e.g. "moderate this", "is this safe", "filter this content"), so it lands at 'some relevant keywords but missing common variations'.	2 / 3
Distinctiveness Conflict Risk	The named model, six specific safety categories, and integration targets carve out a clear niche unlikely to conflict with other skills.	3 / 3
	Total	10 / 12 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: Orchestra-Research/AI-Research-SKILLs
Path: 07-safety-alignment/llamaguard/SKILL.md
Commit: 773a529

Reviewed: about 12 hours 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.