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.

61

Quality

73%

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 ./backend/cli/skills/llm-tools/llamaguard/SKILL.md
SKILL.md
Quality
Evals
Security

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 body is highly actionable with executable code across five workflows, but it is over-long with redundancy, lacks validation feedback loops, and points to three reference files that do not exist while keeping that material inline.

Suggestions

Create the referenced bundle files (references/custom-categories.md, references/benchmarks.md, references/deployment.md) and move the inlined benchmarks, deployment/hardware, and custom-category material into them so SKILL.md stays an overview.

De-duplicate the repeated vLLM init and throughput/latency figures — define them once and reference them — to tighten conciseness toward level 3.

Add explicit validation checkpoints to the deployment and batch-moderation workflows (e.g., verify the model loads, confirm a known-safe/known-unsafe smoke test before serving) to add the feedback loops the workflow_clarity level-3 anchor expects.

DimensionReasoningScore

Conciseness

Mostly efficient action-oriented code, but it could be tightened: the vLLM init snippet recurs across Workflows 3 and 4, throughput/latency figures are restated in multiple sections, and the opening line re-explains the model already covered by the description — fitting the 'mostly efficient but could be tightened' anchor rather than the lean level 3.

2 / 3

Actionability

Provides fully executable, copy-paste-ready code throughout — transformers inference, input/output filtering helpers, vLLM serving, a FastAPI endpoint, and NeMo registration — with specific commands and concrete examples matching the 'fully executable code; copy-paste ready' anchor.

3 / 3

Workflow Clarity

Workflows are clearly sequenced with complete code, but there are no explicit validation checkpoints or validate→fix→retry feedback loops for the deployment/batch recipes, which the level-3 anchor requires; this fits the 'sequence present but checkpoints missing or implicit' anchor.

2 / 3

Progressive Disclosure

References to references/custom-categories.md, references/benchmarks.md, and references/deployment.md are well-signaled and one level deep, but those bundle files do not exist and the corresponding advanced content (benchmarks, deployment, hardware) is inlined in SKILL.md rather than split out, matching the 'content that should be separate is inline' anchor.

2 / 3

Total

9

/

12

Passed

Description

82%

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 rich in natural trigger terms with a distinct niche, but it omits any explicit 'when to use' guidance, capping completeness at 2.

Suggestions

Add an explicit 'Use when ...' clause (e.g., 'Use when moderating LLM inputs/outputs, filtering unsafe content, or when the user mentions content moderation or safety classification') to fully answer the 'when' side and lift completeness to 3.

Lead with the verb/action (third person) rather than the noun phrase 'Meta's 7-8B specialized moderation model' to foreground what the skill does.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'specialized moderation model for LLM input/output filtering', enumerates 6 safety categories, and names deployment/integration targets ('vLLM, HuggingFace, Sagemaker', 'Integrates with NeMo Guardrails'), matching the 'lists multiple specific concrete actions' anchor.

3 / 3

Completeness

Clearly answers 'what does this do' (moderation model for input/output filtering) but provides no 'Use when...' clause or equivalent explicit trigger guidance, so completeness is capped at 2 per the judging guideline.

2 / 3

Trigger Term Quality

Includes natural terms users would say — 'moderation model', 'input/output filtering', 'safety categories', 'content moderation' — giving good coverage of the phrasing a user would reach for, matching the 'good coverage of natural terms' anchor.

3 / 3

Distinctiveness Conflict Risk

Names a specific product ('Meta's 7-8B specialized moderation model') with a distinct niche and concrete deployment context, making it unlikely to trigger for unrelated skills.

3 / 3

Total

11

/

12

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
synthetic-sciences/openscience
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.