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

60%

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 ./backend/cli/skills/llm-tools/llamaguard/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 body is actionable and well-stocked with executable examples, but it is long, time-sensitive, lacks validation checkpoints in its workflows, and keeps all detail inline rather than offloading to reference files.

Suggestions

Move the full per-workflow implementations and hardware requirements into reference files (e.g. workflows.md, deployment.md) and link to them from SKILL.md.

Add validation/verification steps to the batch and API workflows (e.g. confirm output parses as 'safe'/'unsafe' before acting on it).

Relocate version- and date-sensitive details (model versions, 'Latest (2024)', accuracy numbers) to a clearly marked section so the core guidance stays evergreen.

Dimension	Reasoning	Score
Conciseness	The body is mostly practical code with little conceptual fluff, but it runs ~320 lines and embeds time-sensitive details (model versions, 'Latest (2024)', accuracy/throughput figures) outside any deprecated/old-patterns section, which penalizes conciseness.	3 / 5
Actionability	Provides concrete, mostly copy-paste-ready code across input/output filtering, vLLM, FastAPI, and NeMo workflows, with minor gaps such as the vLLM and API snippets referencing an undefined `tokenizer`.	4 / 5
Workflow Clarity	Workflows are sequenced (install → load → moderate → check result) but lack explicit validation checkpoints or feedback loops, and the batch-moderation workflow has no result verification, capping clarity at 3.	3 / 5
Progressive Disclosure	Section structure is clear, but with no bundle files everything is inlined in one ~320-line SKILL.md; content that belongs in separate references (full workflow implementations, hardware specs) is not split out or signaled.	3 / 5
	Total	13 / 20 Passed

Description

62%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 is specific and distinct, naming a concrete product and its capabilities, but it omits any explicit 'when to use' trigger guidance and lacks some natural keyword variations users would actually say.

Suggestions

Add a 'Use when...' clause stating when Claude should reach for this skill, e.g. moderating LLM inputs/outputs or blocking unsafe content.

Include natural user-facing trigger phrases like 'content moderation', 'toxic content', or 'block harmful prompts'.

Lead with the action Claude performs (classify/filter content) rather than product attributes (parameter count, accuracy).

Dimension	Reasoning	Score
Specificity	Lists several concrete specifics — 'input/output filtering', deployment with vLLM/HuggingFace/Sagemaker, and NeMo Guardrails integration — beyond merely naming the domain, though several read as product attributes rather than actions Claude performs.	4 / 5
Completeness	Clearly states what the model does, but provides no 'Use when...' clause or equivalent trigger guidance, capping completeness at 3 per the missing-trigger guideline.	3 / 5
Trigger Term Quality	Contains relevant keywords ('moderation model', 'input/output filtering', 'safety categories') but misses common natural variations a user might say such as 'content moderation', 'block harmful content', or 'toxic content'.	3 / 5
Distinctiveness Conflict Risk	Names a specific, branded product with a narrow function (Meta's 7-8B moderation model), giving it a clear niche with minimal overlap risk against other skills.	5 / 5
	Total	15 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/llm-tools/llamaguard/SKILL.md
Commit: 7dabbf0

Reviewed: 2 days 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.