CtrlK
BlogDocsLog inGet started
Tessl Logo

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.

56

Quality

65%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./skills/safety-alignment/llamaguard/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

65%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 skill is highly actionable with concrete code for the main moderation workflows, but its organization is weakened by dangling references to missing bundle files and a lack of explicit validation checkpoints in the workflows.

Suggestions

Create the referenced files (references/custom-categories.md, benchmarks.md, deployment.md) or remove the dead links so progressive disclosure points to real material.

Add explicit validation/verification steps to the batch and API workflows (e.g. assert the model output parses as 'safe'/'unsafe' before acting on it, with a retry/fallback on parse failure).

Tighten repeated content — the model intro and moderate() setup appear in multiple sections — to improve token efficiency.

DimensionReasoningScore

Conciseness

The body is mostly dense, executable code with little concept padding, but repeats the model intro from the frontmatter and could trim some redundant commentary.

4 / 5

Actionability

Provides copy-paste-ready code across the common cases (input/output filtering, vLLM, FastAPI, NeMo), with minor gaps such as an undefined tokenizer in the vLLM/API examples.

4 / 5

Workflow Clarity

Each workflow is a clear sequenced procedure with branching, but there are no explicit validation checkpoints or validate-fix-retry feedback loops; the troubleshooting section partly compensates.

3 / 5

Progressive Disclosure

Section structure is reasonable and advanced topics are pushed to references, but the three referenced files (custom-categories.md, benchmarks.md, deployment.md) do not exist, and substantial workflow content stays inline.

3 / 5

Total

14

/

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 is specific and concrete about capabilities and deployment, with a clear niche. Its main weakness is the absence of an explicit 'when to use' trigger clause, which caps completeness.

Suggestions

Add an explicit 'Use when...' clause, e.g. 'Use when moderating LLM input/output, filtering unsafe prompts or responses, or when the user mentions content moderation, safety classification, or LlamaGuard.'

Include a few more natural trigger synonyms such as 'toxicity', 'flag unsafe content', or 'block harmful prompts' to broaden keyword coverage.

Tighten the accuracy/deployment list into the trigger phrasing so the 'when' is as concrete as the 'what'.

DimensionReasoningScore

Specificity

Lists several concrete capabilities — 'input/output filtering', six enumerated safety categories, and deployment targets (vLLM, HuggingFace, Sagemaker) plus NeMo Guardrails integration — but stops short of a comprehensive action set.

4 / 5

Completeness

The 'what' is clear (a moderation model for input/output filtering), but there is no 'Use when...' clause or equivalent explicit trigger guidance, capping completeness at 3 per the rubric.

3 / 5

Trigger Term Quality

Includes natural terms like 'moderation', 'filtering', and 'safety categories' a user would say, but misses common synonyms such as 'toxicity', 'block unsafe', or 'flag content'.

4 / 5

Distinctiveness Conflict Risk

Naming Meta's specific LlamaGuard model and its six categories gives it a clear niche with minimal conflict risk, though 'safety/moderation' broadly could overlap slightly with general safety skills.

4 / 5

Total

15

/

20

Passed

Validation

81%

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

Validation13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

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
Reviewed

Table of Contents

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.