skills.rest
DeepLink-org
DeepLink-orgOFFICIAL
Explore all skills by this author →
WebsiteGitHub

template-skill

Provides a reusable template for authoring hardware benchmark evaluation skills.

7|1|Updated Apr 17, 2026
Search Tags:#template#agent skills#docker#skill authoring#benchmark#performance testing#gpu evaluation
One-click install
npx skills add https://github.com/DeepLink-org/DeepEval-Skills --skill template-skill-deeplink-org
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: template-skill
Source: https://github.com/DeepLink-org/DeepEval-Skills/tree/main/template
Command: npx skills add https://github.com/DeepLink-org/DeepEval-Skills --skill template-skill-deeplink-org

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? This Skill gives benchmark engineers a standardized starting point for creating new AI chip evaluation skills, eliminating the guesswork of structuring SKILL.md files that comply with the Agent Skills specification. ## Core Features & Use Cases - Standardized Frontmatter: Includes pre-defined YAML fields for name, description, compatibility, category, scenario, and tags used for skill discovery. - Complete Evaluation Skeleton: Provides ready-made sections for hardware requirements, Docker container startup, environment variables, execution steps, and performance metric tables. - Use Case: When adding a new benchmark (e.g., evaluating GEMM operator performance on a new GPU platform), copy this template, fill in the chip platform, model, dataset paths, and metric definitions, and the skill is ready for registration. ## Quick Start Copy this template directory and fill in the chip platform, model, dataset, and performance metric sections to create a new benchmark evaluation skill.

Frequently Asked Questions about template-skill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a new benchmark evaluation skill?▼

Copy the template directory, then edit SKILL.md to replace the placeholder frontmatter (name, description, compatibility, category, scenario) and fill in the hardware requirements, Docker image, environment variables, and execution steps for your specific benchmark.

What fields are required in a SKILL.md frontmatter?▼

The mandatory fields are name and description, which are used for skill discovery. Optional fields include user-invokable, compatibility, and a metadata block with version, category, scenario, and tags.

What categories and scenarios does the template support?▼

The template metadata supports four categories: training, finetune, inference, and operator. Supported scenarios include nlp, cv, mm (multimodal), science, and audio.

Which chip platforms can benchmark skills target?▼

The compatibility field accepts platforms such as NVIDIA GPU, Hygon DCU, or other accelerator chips. The template's Docker-based workflow uses GPU passthrough flags that can be adapted per platform.

What performance metrics should a benchmark skill collect?▼

The template recommends tokens_per_sec_per_gpu as a mandatory performance metric, with GPU utilization and memory usage as auxiliary metrics, and accuracy metrics like RMSE or ACC as optional precision measurements.

Related Skills

Explore More Software Engineering Skills→
EurecaMomentEurecaMoment

benchclaw-stage3-real-image-cleaning

Automate cleaning of real-image datasets for BenchClaw benchmarking with Data-Juicer.

Official
Intermediate
howdeployhowdeploy
15

benchmark

Benchmark web pages against baselines to detect performance regressions.

Community
Advanced
NVIDIANVIDIA
14.5k

exec-local-docker

Executes TensorRT-LLM workloads in local Docker containers with background monitoring and log-based result reporting.

Official
Advanced
NVIDIANVIDIA
3.2k

jetson-init-target

Author Jetson target-platform YAML profiles with reference devkit and optional custom carrier details.

Official
Advanced