skill-test（技能测试） Skills下载安装命令和教程

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手，先确认命令和权限，再让它执行。

请帮我安装这个 Agent Skill：skill-test（技能测试）
来源仓库：https://github.com/microsoft/skills-for-fabric
仓库路径：skills/skill-test
安装命令：
npx skills add https://github.com/microsoft/skills-for-fabric --skill skill-test
安装前请先检查当前环境是否支持对应 CLI，并向我确认将要执行的命令、安装目录、联网范围和文件读写权限；确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill；本站只展示命令，不托管安装包，也不自动执行。

skills.shnpx skills

npx skills add https://github.com/microsoft/skills-for-fabric --skill skill-test

简介

用于辅助测试设计、用例整理和回归验证。skill-test 属于待分类类 Skill，可作为该场景下的辅助能力补充。

适合编写单元测试、端到端测试或根据日志定位问题。
需确认项目测试框架和运行命令，避免误改逻辑。
涉及浏览器或服务调用时应区分模拟与真实环境。
安装方式：通过 GitHub 仓库使用 npx 命令添加技能。

SKILL.md

Skill Test — skills-for-fabric Evaluation Framework

Manage the end-to-end evaluation framework for skills-for-fabric. This skill routes requests to the correct workflow based on user intent — adding tests, listing tests, running tests, viewing results, generating data, or checking coverage.

When to Use

When a contributor wants to add evaluation test cases for a new or existing skill
When someone asks to see what tests exist or what results look like
When a user wants to run the test suite
When reviewing eval metrics or checking which skills lack test coverage

Intent Routing

Parse the user request and route to the appropriate workflow:

User Intent	Trigger Phrases	Action
Add evals	"add tests", "add evals", "add evals for missing skills", "create eval plan"	→ Workflow: Add Evals
List tests	"list tests", "list evals", "show me the list of tests", "what tests exist", "show eval plans"	→ Workflow: List Tests
Run tests	"run tests", "run evals", "execute tests", "run the eval suite"	→ Workflow: Run Tests
View results	"show eval results", "test results", "eval results", "executive summary"	→ Workflow: View Results
Generate data	"generate eval data", "generate test data", "create eval datasets"	→ Workflow: Generate Data
View metrics	"eval metrics", "test metrics", "what metrics", "how are tests scored"	→ Workflow: View Metrics
Check coverage	"test coverage", "which skills have tests", "missing tests", "skills without evals"	→ Workflow: Check Coverage

Workflow: Add Evals

Follow the instructions in tests/full-eval-tests/README.md § "Adding Evals for New Skills".

Automated Path (Recommended)

Give the agent the prompt:

Add evals for the missing skills

The agent will:

Detect missing skills by comparing installed skills against existing eval plans in tests/full-eval-tests/plan/03-individual-skills/
Generate individual eval plans (plan/03-individual-skills/eval-<skill-name>.md) with 10–12 test cases
Generate combined eval plans (plan/04-combined-skills/eval-<skill>-authoring-plus-consumption.md)
Create golden data in tests/full-eval-tests/evalsets/expected-results/
Update tracking files: plan/00-overview.md, README.md, plan/04-combined-skills/eval-full-pipeline.md

Manual Path

To add evals for a specific skill <new-skill>:

Create tests/full-eval-tests/plan/03-individual-skills/eval-<new-skill>.md using the template in the README
Each test case needs: Case ID (unique prefix), Prompt, Expected result, Pass criteria, at least one negative/ambiguous test
If the skill has an authoring+consumption pair, create tests/full-eval-tests/plan/04-combined-skills/eval-<new-skill>-authoring-plus-consumption.md
Add golden data to tests/full-eval-tests/evalsets/expected-results/
Update plan/00-overview.md, README.md directory tree, and plan/04-combined-skills/eval-full-pipeline.md

Eval Plan Template

Use the template from tests/full-eval-tests/README.md § "Eval Plan Template". Every eval plan must include:

Skill overview (name, category, R/W, purpose)
Pre-requisites
Numbered test cases (XX-01 through XX-10+) with Prompt / Expected / Pass criteria
At least one negative/ambiguous test case as the last case
Write Operations table (if the skill writes data)
Expected Token Range

Workflow: List Tests

Show the user what eval plans and test cases exist.

Individual Skill Evals

List files in tests/full-eval-tests/plan/03-individual-skills/:

ls tests/full-eval-tests/plan/03-individual-skills/

Combined Skill Evals

List files in tests/full-eval-tests/plan/04-combined-skills/:

ls tests/full-eval-tests/plan/04-combined-skills/

Quick Tests (tests.json)

Show the test cases defined in tests/tests.json — these are the prompt-based tests run by the test runner.

Recommended Execution Order

Order	Eval Plan	Reason
1	eval-check-updates.md	Verify skills are installed
2	eval-spark-authoring.md	Create lakehouses and load data
3	eval-sqldw-authoring.md	Create warehouse tables and load data
4	eval-eventhouse-authoring.md	Create Eventhouse tables and ingest data
5	eval-spark-consumption.md	Read back lakehouse data
6	eval-sqldw-consumption.md	Read back warehouse data
7	eval-eventhouse-consumption.md	Read back Eventhouse data
8	eval-medallion.md	End-to-end medallion pipeline

Workflow: Run Tests

⛔ DO NOT execute tests from this skill. The agent must NEVER run copilot, run-full-tests.ps1, or any eval prompt directly. Instead, tell the user the exact commands to run manually.

When the user asks to run tests, respond only with instructions. Do not execute any commands. Tell the user:

Open a terminal and navigate to the tests/ directory at the repository root: cd tests
Run the full test suite: .\run-full-tests.ps1
To specify an output directory: .\run-full-tests.ps1 -TestFolder C:\temp\eval-run-01

Important

The agent must NEVER run tests itself — only provide the user with instructions
Tests must be run by the user from inside the tests/ folder
The script copies the eval framework to a working folder and launches copilot there

Workflow: View Results

Show the user existing evaluation results.

Detailed Results

Read tests/full-eval-tests/eval-results.md — contains per-skill, per-test-case pass/fail with notes, consistency test results, failure analysis, and skip reasons.

Executive Summary

Read tests/full-eval-tests/executive-summary.md — contains the high-level summary: overall pass rate, results by skill, data consistency scores, failure analysis, and recommendations.

Key Metrics from Latest Run

Metric	Value
Overall pass rate	94.7% (54/57 executed)
Write/Read consistency	100% (5/5 exact matches)
Total test cases	74
Skipped	17

Workflow: Generate Data

Generate synthetic evaluation datasets using the specifications in tests/full-eval-tests/plan/01-data-generation.md.

Using the Generation Script

python tests/full-eval-tests/evalsets/data-generation/generate.py

Datasets

Dataset	Rows	Format	Used By
sales_transactions	100 / 1K / 10K	CSV	SQL DW, Spark
customers	100	CSV	Join testing
products	50	CSV	Join testing
sensor_readings	500	JSON	Spark semi-structured

Golden Results

Pre-computed expected results are in tests/full-eval-tests/evalsets/expected-results/ and are used to verify consistency.

Workflow: View Metrics

Explain the evaluation metrics defined in tests/full-eval-tests/plan/02-metrics.md.

Metric	Definition
Success Rate	`passed / total × 100` — whether the skill executed correctly
Token Usage	Input + output tokens consumed per eval prompt
Read/Write Consistency	Data written by authoring skill must be exactly retrievable by consumption skill

Grading

Grade	Criteria
PASS	Skill invoked correctly, output matches expected
FAIL_INVOCATION	Wrong skill invoked or not invoked
FAIL_EXECUTION	Skill invoked but errored
FAIL_RESULT	Skill completed but output mismatches

Pass Thresholds

Metric	Threshold
Success Rate	≥ 90% per skill
Token Usage	Within 2× of baseline
Read/Write Consistency	100% exact match

Workflow: Check Coverage

Compare installed skills against existing eval plans to identify gaps.

Steps

List all skills from the marketplace/plugin: check-updates, spark-authoring-cli, spark-consumption-cli, sqldw-authoring-cli, sqldw-consumption-cli, eventhouse-authoring-cli, eventhouse-consumption-cli, e2e-medallion-architecture
List existing individual eval plans: ls tests/full-eval-tests/plan/03-individual-skills/
Compare and report which skills have eval coverage and which are missing.
For missing skills, suggest running the Add Evals workflow.

Must

NEVER execute tests, eval prompts, or the test runner script — only provide instructions for the user to run manually
Always route "run tests" to the tests/ folder — tell the user to navigate there and run run-full-tests.ps1
Follow the eval plan template when creating new eval plans — every test case needs Case ID, Prompt, Expected, Pass criteria
Include at least one negative/ambiguous test in every new eval plan
Update tracking files when adding evals — plan/00-overview.md, README.md, eval-full-pipeline.md
Reference golden data in evalsets/expected-results/ for consistency tests

Prefer

Automated eval generation ("add evals for missing skills") over manual creation
Reading existing eval plans as templates before creating new ones
Running the full suite rather than individual tests for comprehensive coverage
Checking the executive summary before diving into detailed results

Avoid

Running eval prompts outside the tests/ directory
Executing tests, copilot commands, or the run-full-tests.ps1 script — the agent must only tell the user how to run them
Creating eval plans without a negative/ambiguous test case
Duplicating content already in the README or plan documents — reference them instead
Modifying golden result files unless the underlying data generation rules change
Skipping the tracking file updates when adding new evals

Examples

Adding Tests

User: "Add tests for the powerbi-consumption-cli skill"

Agent: Creates tests/full-eval-tests/plan/03-individual-skills/eval-powerbi-consumption.md with 10–12 test cases following the template, adds golden data to evalsets/expected-results/, and updates tracking files.

Listing Tests

User: "What tests exist?"

Agent: Lists all eval plans in plan/03-individual-skills/ and plan/04-combined-skills/, plus the quick tests in tests/tests.json, with the recommended execution order.

Running Tests

User: "Run the tests"

Agent:

Tests must be run from the tests/ folder. Navigate there and execute: ``powershell cd tests .\run-full-tests.ps1 ``

Viewing Results

User: "Show me the eval results"

Agent: Reads tests/full-eval-tests/eval-results.md and presents the summary table, highlighting pass rates and any failures.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明，方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用；具体调用方式请以来源页面和 README 为准。

平台分布

Codex

37.02%

按下载量换算55

Claude

28.12%

按下载量换算42

Cursor

20.32%

按下载量换算30

Gemini CLI

9.18%

按下载量换算14

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行，可能通过 npx skills add https://github.com/microsoft/skills-for-fabric --skill skill-test 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息，不托管安装包，不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源，正式发布前建议补源仓库或其他目录站核验。

来源信息

GitHub 仓库

github.com

/microsoft/skills-for-fabric

SKILL.md 来源

github.com

/microsoft/skills-for-fabric/tree/main/skills/skill-test

skills-sh

skills.sh

/microsoft/skills-for-fabric/skill-test

继续浏览同类 Skills

更多待分类 Skills（11,209）返回 Skills 目录

上一个 Skill

skill-test（技能测试）

用于辅助数据整理、表格处理、CSV/Excel 分析、指标计算和图表准备。它适合让 Agent 清洗字段、汇总数据、发现异常、生成统计口径或把分析结果转成可读说明。使用时需要确认数据来源、字段含义和时间范围，避免把样本数据当全量事实；涉及敏感数据、导出文件或批量写回时，应先确认权限和脱敏边界。

下一个 Skill

skill-usage（技能运用）

skill-usage 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息，适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态，以及是否会触发联网、命令执行或文件读写。