sglang

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

Quality

89%

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

Quality

Content

86%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.

A comprehensive, highly actionable skill body with strong progressive disclosure into real reference files. The main weakness is minor redundancy between the performance prose and benchmark tables that could be tightened.

Suggestions

Consolidate the "Performance benchmarks" section — the throughput table and "RadixAttention speedup" bullets restate figures already in the "RadixAttention" and "When to use" sections; keep one authoritative source.

For multi-GPU / speculative-decoding server launches, add a one-line verification step (e.g. curl the /v1/models endpoint) so users confirm the server is healthy before running inference.

The grammar-based generation example uses a simplified EBNF that will not produce valid Python in all cases; note it as illustrative or link to a complete grammar reference.

Dimension	Reasoning	Score
Conciseness	Mostly code-forward and efficient, but benchmark tables and the standalone "Performance" bullet restate speeds already conveyed in prose, which could be trimmed; not enough padding to drop to 3.	4 / 5
Actionability	Copy-paste-ready code and commands span installation, server launch, inference, JSON/regex/grammar constraints, agents, multi-turn chat, multi-modal, and the OpenAI API — fully executable with common cases covered.	5 / 5
Workflow Clarity	Sections are clearly sequenced (install → launch → infer → patterns → advanced), and the task type (run server, call API) does not require destructive/batch validation feedback loops, but no explicit checkpoints or checklists are present, capping it below 5.	4 / 5
Progressive Disclosure	SKILL.md serves as a concise overview with three well-signaled, one-level-deep references (structured-generation.md, radix-attention.md, deployment.md) that are real files, with bulk detail pushed to them — easy to navigate.	5 / 5
	Total	18 / 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.

A strong, specific description that clearly states capabilities and explicit use-conditions with natural trigger terms. Only minor room for added synonym coverage in trigger terms.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — "structured generation", "serving for LLMs", "JSON/regex outputs", "constrained decoding", "agentic workflows with tool calls", "prefix sharing" — giving comprehensive coverage rather than the minor gaps of a 4.	5 / 5
Completeness	Explicitly answers both what ("Fast structured generation and serving for LLMs with RadixAttention prefix caching") and when ("Use for JSON/regex outputs, constrained decoding, agentic workflows...") with concrete trigger phrases.	5 / 5
Trigger Term Quality	Good natural keyword coverage ("JSON/regex outputs", "constrained decoding", "agentic workflows with tool calls", "5× faster inference than vLLM"), but a few synonyms/extensions a user might naturally say are missing, so it stops short of the comprehensive 5.	4 / 5
Distinctiveness Conflict Risk	Clear niche (SGLang structured generation with RadixAttention) and a named differentiator against vLLM, giving minimal conflict risk with other skills.	5 / 5
	Total	19 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/ml-inference/sglang/SKILL.md
Commit: 4e060d6

Reviewed: 7 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.