minicpm5-deploy-sglang

Serve MiniCPM5-1B 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

91%

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

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.

A tight, highly actionable deploy runbook with executable commands and a clear numbered workflow including validation. Minor gaps: time-sensitive version info is not segregated and the validation step lacks an explicit error-recovery loop.

Suggestions

Move version/date-sensitive details (PR #25600 merge date, v0.5.12.post1, CUDA driver versions) into a dedicated 'Version notes' or 'Deprecated/old patterns' section so stale info is easy to spot and prune.

Add a short feedback loop after the Validate step, e.g. 'If the response does not contain "2", check server logs and confirm --tool-call-parser and --model-path are set correctly before retrying.'

Consider splitting the Tool calling and Offline/batched examples into a referenced file if the runbook grows, keeping SKILL.md as the launch overview.

Dimension	Reasoning	Score
Conciseness	Lean runbook with tables and code blocks that assumes competence, but time-sensitive version/date details ('PR #25600, merged 2026-05-22', 'v0.5.12.post1', 'CUDA 13.x') are not isolated in a deprecated/old-patterns section, which the rubric penalizes.	4 / 5
Actionability	Fully executable, copy-paste-ready install commands, launch invocation with all flags, validation curl with expected output, a tool-calling request example, and a batch Engine API snippet covering the common cases.	5 / 5
Workflow Clarity	A clearly numbered 1-4 sequence with a 'Wait for server fired up' readiness check and a validation step with expected output, but it lacks an explicit validate->fix->retry feedback loop for when validation fails.	4 / 5
Progressive Disclosure	Well-organized single-file runbook with clear section headers and one well-signaled one-level-deep reference (docs/deployment/sglang.md); appropriate for a deploy runbook though most content is inlined rather than split out.	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.

A strong, concise description that names concrete capabilities, uses third person, and provides an explicit trigger clause with natural synonyms. It is clearly distinguishable from sibling deploy skills.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — serving via SGLang as an OpenAI-compatible HTTP server, RadixAttention prefix cache, and built-in MiniCPM5 tool-call parsing — giving comprehensive coverage of the skill's capabilities.	5 / 5
Completeness	Explicitly answers both 'what' (serve MiniCPM5-1B via SGLang with prefix cache and tool-call parsing) and 'when' via a clear 'Use when the user asks for...' clause with concrete trigger phrases.	5 / 5
Trigger Term Quality	Comprehensive natural trigger terms including 'SGLang', 'RadixAttention', 'prefix cache', 'batch evaluation', 'tool calling', and the synonym/alternative 'high-concurrency NVIDIA-GPU server alternative to vLLM' that users would naturally say.	5 / 5
Distinctiveness Conflict Risk	Clear niche (SGLang serving of MiniCPM5 specifically) with distinct triggers and an explicit vLLM alternative, making overlap with other deploy skills minimal.	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: 719e4fc

Reviewed: 17 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.