minicpm5-deploy-sglang

Serve MiniCPM5-1B or MiniCPM5-2B via SGLang as an OpenAI-compatible HTTP server with RadixAttention prefix cache and built-in MiniCPM5 tool-call parsing. Use when the user asks for "SGLang", "RadixAttention", "prefix cache", batch evaluation, tool calling, or wants a high-concurrency NVIDIA-GPU server alternative to vLLM.

Quality

92%

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

Quality

Content

82%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 highly actionable with a clear, validated deployment workflow and tight prose. The main gaps are inline time-sensitive version details, a missing error-recovery feedback loop, and a non-resolving external reference.

Suggestions

Move time-sensitive version/date details (PR #25600, the 2026-05-22 merge date, v0.5.12.post1) into a dedicated "Version notes" or "Deprecated/old patterns" section so the main install flow stays lean.

Add an explicit validate→fix→retry feedback loop after the Validate step (e.g., if the curl does not return "2", check server logs / confirm the minicpm5 parser is installed from main, then relaunch) to lift workflow clarity.

Make the "Reference" link reliable — either confirm docs/deployment/sglang.md exists in the bundle, inline the key details, or remove the broken pointer.

Dimension	Reasoning	Score
Conciseness	The body is mostly lean and assumes Claude's competence, but time-sensitive specifics ("PR #25600, merged 2026-05-22", "v0.5.12.post1", ">=0.5.16") sit inline in the main install flow rather than a version/deprecated section, and the speculative-decoding block repeats the full launch command — minor trimming opportunities keep it just below 5.	4 / 5
Actionability	Fully executable, copy-paste-ready guidance across the common cases: pip install with a fallback, env-var exports, a complete launch command, a validate curl with expected output, a tool-calling curl, and an offline Engine snippet.	5 / 5
Workflow Clarity	A clear numbered sequence (Install → env vars → Launch → Validate) with an explicit readiness checkpoint ("The server is fired up and ready to roll!") and an expected-output check, but there is no explicit validate→fix→retry feedback loop, so it sits at 4 rather than 5.	4 / 5
Progressive Disclosure	Well-organized sections with a single one-level-deep Reference pointer, but the skill is ~145 lines (over the simple-skill 50-line threshold) with most content inline, and the sole external reference (../../docs/deployment/sglang.md) does not resolve in the bundle — good structure with minor organization gaps.	4 / 5
	Total	17 / 20 Passed

Description

100%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, trigger-rich, and fully answers both what and when in third person. It is a strong, activation-driving description with no significant gaps.

Dimension	Reasoning	Score
Specificity	"Serve MiniCPM5-1B or MiniCPM5-2B via SGLang as an OpenAI-compatible HTTP server with RadixAttention prefix cache and built-in MiniCPM5 tool-call parsing" lists multiple concrete actions (serve model, expose OpenAI-compatible HTTP, prefix cache, tool-call parsing) with comprehensive coverage, matching the 5 anchor.	5 / 5
Completeness	It explicitly answers "what" (serve via SGLang as OpenAI-compatible server with prefix cache and tool-call parsing) and "when" ("Use when the user asks for...") with concrete trigger phrases, matching the 5 anchor.	5 / 5
Trigger Term Quality	Trigger phrases "SGLang", "RadixAttention", "prefix cache", "batch evaluation", "tool calling", and "high-concurrency NVIDIA-GPU server alternative to vLLM" are exactly what a user would say, with synonyms/variants (vLLM alternative, high-concurrency serving) — comprehensive natural-term coverage.	5 / 5
Distinctiveness Conflict Risk	The SGLang + MiniCPM5 niche is clearly scoped with distinct triggers and an explicit contrast to vLLM, giving minimal overlap risk with sibling deploy skills.	5 / 5
	Total	20 / 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
relative_links	Relative link issues: 1 suspicious	Warning

	Total	15 / 16 Passed

Repository: OpenBMB/MiniCPM
Path: skills/minicpm5-deploy-sglang/SKILL.md
Commit: 316cfb1

Reviewed: 15 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.