dokimos-dev/dokimos/create-experiment

Claude Code and agent skill by dokimos-dev/dokimos

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Scaffolds a Dokimos Experiment that wires together a dataset, task, evaluators, and optional reporter. Use this skill when the user wants to create an evaluation experiment, set up an eval pipeline, run evaluations against a dataset, or build an end-to-end evaluation workflow. Also use when the user says "evaluate my LLM" or "test my model".

Install

npx vskill@latest install dokimos-dev/dokimos/create-experiment

SKILL.md

Create Experiment

Scaffold a Dokimos Experiment. The user will describe the evaluation goal via $ARGUMENTS.

Where things live

  • Experiment class: dokimos-core/src/main/java/dev/dokimos/core/Experiment.java
  • Task interface: dokimos-core/src/main/java/dev/dokimos/core/Task.java
  • Example experiments: dokimos-examples/src/main/java/dev/dokimos/examples/

Before writing code, read these files to understand the API:

  • Experiment.java — the orchestrator
  • dokimos-examples/src/main/java/dev/dokimos/examples/basic/BasicEvaluationExample.java — simplest example

Experiment anatomy

An experiment consists of four parts:

  1. Dataset — the test cases (see create-dataset skill)
  2. Task — a function that takes an Example and returns Map<String, Object> of actual outputs
  3. Evaluators — one or more Evaluator implementations that score the outputs
  4. Reporter (optional) — sends results to the Dokimos server or elsewhere
Basic template
java
Dataset dataset = Dataset.fromJson(Path.of("src/test/resources/datasets/my-dataset.json"));
Task task = example -> {
String input = example.input();
String output = callYourLLM(input);
return Map.of("output", output);
};
List<Evaluator> evaluators = List.of(
ExactMatchEvaluator.builder()
.name("Exact Match")
.threshold(1.0)
.build()
);
ExperimentResult result = Experiment.builder()
.name("My Evaluation")
.description("Evaluating my LLM on QA tasks")
.dataset(dataset)
.task(task)
.evaluators(evaluators)
.build()
.run();
System.out.println("Pass rate: " + result.passRate());
With JUnit integration
java
public class MyEvaluationTest {
@ParameterizedTest(name = "[{index}] {0}")
@DatasetSource("classpath:datasets/my-dataset.json")
void testMyLLM(Example example) {
String actualOutput = callYourLLM(example.input());
EvalTestCase testCase = example.toTestCase(actualOutput);
List<Evaluator> evaluators = List.of(
ExactMatchEvaluator.builder().build()
);
assertEval(testCase, evaluators);
}
}
With server reporting
java
Reporter reporter = DokimosServerReporter.builder()
.baseUrl("http://localhost:8080")
.build();
ExperimentResult result = Experiment.builder()
.name("My Evaluation")
.dataset(dataset)
.task(task)
.evaluators(evaluators)
.reporter(reporter)
.build()
.run();

Builder options

MethodDescriptionDefault
.name(String)Experiment name"unnamed"
.description(String)Description""
.dataset(Dataset)Test datasetrequired
.task(Task)System under testrequired
.evaluator(Evaluator)Add a single evaluator—
.evaluators(List)Add multiple evaluators—
.reporter(Reporter)Result reporterNoOpReporter
.parallelism(int)Concurrent examples1
.runs(int)Repeat count for variance reduction1
.metadata(String, Object)Add experiment metadata—

Steps

  1. Determine from $ARGUMENTS what the user wants to evaluate
  2. Choose the right integration (plain Java, LangChain4j, Spring AI, Koog)
  3. Create or reference a dataset
  4. Define the task function wrapping the user's system
  5. Select appropriate evaluators for the use case
  6. Wire everything together with Experiment.builder()
  7. Add a JUnit test if the experiment should run in CI

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
42
Repository forks
4
7-day trend score
19

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/dokimos-dev/dokimos/create-experiment/badge)](https://verified-skill.com/skills/dokimos-dev/dokimos/create-experiment)

Something wrong with this skill or its page? Tell us.