CtrlK
BlogDocsLog inGet started
Tessl Logo

serving-llms-vllm

Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.

65

Quality

78%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-inference/vllm/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 body is well-structured with executable examples and excellent progressive disclosure via real reference files. Its main weaknesses are verbose filler and missing validation/verification checkpoints in batch and deployment workflows.

Suggestions

Add explicit validation/verification steps to the batch inference workflow (e.g., assert output count matches input count, validate JSONL is well-formed) to raise workflow clarity.

Remove explanatory filler such as 'vLLM handles batching internally / No need to manually chunk prompts' and replace comment-only placeholders like the quantization accuracy check with concrete runnable checks.

Tighten the deployment checklist steps so each step has a concrete verification command rather than generic 'Verify performance metrics' guidance.

DimensionReasoningScore

Conciseness

Mostly efficient with tight code blocks, but includes filler like 'vLLM handles batching internally / No need to manually chunk prompts' and placeholder comment-only blocks ('# Compare quantized vs non-quantized responses') that could be trimmed.

3 / 5

Actionability

Provides mostly executable, copy-paste-ready commands and code across all workflows, with only minor gaps such as the quantization 'Verify accuracy' step being comments rather than runnable code.

4 / 5

Workflow Clarity

Workflows are sequenced with checklists, but batch/deploy operations lack explicit validation checkpoints (e.g., the batch workflow never verifies output integrity), capping the score per the rubric.

3 / 5

Progressive Disclosure

Body is a clear overview with well-signaled, one-level-deep references to real bundle files (server-deployment.md, optimization.md, quantization.md, troubleshooting.md), all of which exist in ./references/.

5 / 5

Total

15

/

20

Passed

Description

92%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 strong, specific, and complete with explicit trigger guidance and concrete capability lists. It clearly conveys both what the skill does and when to use it, with minimal conflict risk.

DimensionReasoningScore

Specificity

Lists multiple concrete capabilities — 'PagedAttention and continuous batching', 'OpenAI-compatible endpoints', 'quantization (GPTQ/AWQ/FP8)', 'tensor parallelism' — giving comprehensive coverage of what the skill does.

5 / 5

Completeness

Clearly answers both 'what' (serves LLMs with high throughput via PagedAttention/continuous batching) and 'when' with an explicit 'Use when deploying production LLM APIs...' trigger clause.

5 / 5

Trigger Term Quality

Includes good natural triggers like 'deploying production LLM APIs', 'optimizing inference latency/throughput', and 'serving models with limited GPU memory', but lacks common synonyms or extension-style phrasings a user might naturally say.

4 / 5

Distinctiveness Conflict Risk

Occupies a clear vLLM-specific serving niche with distinct triggers; named technologies (PagedAttention, FP8/GPTQ/AWQ) keep it from overlapping with generic inference skills.

5 / 5

Total

19

/

20

Passed

Validation

93%

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

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

15

/

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.