Skip to content

aaravarr/test-creator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

test-creator

Stop guessing what to test. Give your project a professional, production-grade test suite — with one sentence.

License: MIT Language Agnostic Framework Agnostic

English | 中文


test-creator is a skill that builds a complete, professional test system for any project. It doesn't fix your code — it tells you exactly what's broken and where.

Works with any language, any framework — Go, Python, Node.js, Java, Rust, you name it.


What It Does

test-creator gives your AI coding agent a structured, battle-tested methodology for building test suites. It's not a code generator — it's a thinking framework with automated quality tooling that ensures nothing falls through the cracks.

The 6-Point Coverage Guarantee

Every test type must pass all 6 mandatory checkpoints. No shortcuts.

Checkpoint Why It Matters
1 Happy Path Does it actually work?
2 Boundaries What happens with empty inputs, max values, special characters?
3 Status Codes Are you getting the right HTTP/business codes — not just 200?
4 Data Integrity Does the database actually store what was sent? Field-by-field verification against the storage layer.
5 Logging Do critical operations log properly? Is sensitive data staying out of logs?
6 Error Handling What happens when things go wrong? Graceful degradation or crash?

4 Test Types

Pick what your project needs — or use all four:

  • API Testing — CRUD, auth, pagination, concurrency, idempotency
  • E2E Testing — User flows, forms, routing, async state
  • Unit Testing — Pure functions, transforms, business rules, state machines
  • Integration Testing — Database consistency, 3rd-party services, message queues

Dual-Layer Quality Assurance

Layer 1: Automated Scripts — Coverage, stability (flaky test detection), and performance are measured mechanically. Three dimensions, one report.

Layer 2: Adversarial Sub-Agent Review — A separate AI agent reviews the test code from 6 dimensions, finding issues scripts can't: incorrect assertions, missing scenarios, unreasonable mocks, poor test data. The reviewer's job is to find problems, not validate good work.


Why test-creator?

The old way With test-creator
Ask the agent "write tests" and hope for the best 7-step workflow: deep analysis → Q&A → test plan → implement → verify → review → fix
Tests pass but miss real bugs 6 mandatory test points per type — no cherry-picking
Data validation = checking response field types Data validation = querying the storage layer directly, field by field
No idea if the suite is actually good Multi-dimensional quality report (coverage + stability + performance)
Self-review (agent checking its own work) Adversarial sub-agent review with 6 dimensions
Works with one framework Works with any language/framework via adapter system

Supported Frameworks

Auto-detected — zero configuration needed:

  • JavaScript/TypeScript — Jest, Vitest
  • Python — pytest
  • Go — go test
  • Java — JUnit

Don't see your framework? The adapter system is extensible — add your own in under 50 lines of shell script.


Installation

Claude Code

curl -sSL https://raw.githubusercontent.com/ahaostudy/test-creator/main/scripts/install.sh | bash -s -- --tool claude-code

Codex

curl -sSL https://raw.githubusercontent.com/ahaostudy/test-creator/main/scripts/install.sh | bash -s -- --tool codex

Installs to ~/.agents/skills/test-creator/. Use /skills or $ to invoke.

OpenClaw

curl -sSL https://raw.githubusercontent.com/ahaostudy/test-creator/main/scripts/install.sh | bash -s -- --tool openclaw

Installs to ~/.openclaw/skills/test-creator/.

Manual / Any Tool

curl -sSL https://raw.githubusercontent.com/ahaostudy/test-creator/main/scripts/install.sh | bash -s -- --tool generic --dir /your/target/dir

Or just copy the directory yourself — the skill needs SKILL.md plus the adapters/, scripts/, and references/ folders alongside it.

Quick Start

1. Install

Pick your tool above and run the install command.

2. Use it

Ask your AI agent to build tests for any project:

"Help me create a test suite for my Flask API"

"I want to add integration tests to this Go project"

"Build E2E tests for this React app"

test-creator activates automatically.

3. Let it work

Your agent will:

  1. Run parallel sub-agents to deeply analyze your project — every API endpoint, page flow, data model, and log config
  2. Generate an interactive Q&A page for you to configure what to cover
  3. Write a test development plan and confirm it with you before writing any code
  4. Implement the test suite, all files organized under tests/
  5. Run automated quality checks via run-all-checks.sh across 3 dimensions
  6. Have a sub-agent review the tests from 6 angles
  7. Deliver a quality report showing exactly what passes, what fails, and why

License

MIT

About

test-creator is a skill that builds a complete, professional test system for any project. It doesn't fix your code — it tells you exactly what's broken and where.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Contributors

Languages