QAPI Test Creation Skill

This repository contains a reusable AI skill that scans a codebase for HTTP API endpoints, builds QAPI-format test payloads, and saves them to a QAPI workspace through the bundled Python client.

Repository Contents

qapi-skill/
├── SKILL.md                # Core agent instructions
├── agents/openai.yaml      # Optional UI metadata for skill-capable clients
├── scripts/qapi_client.py  # QAPI REST client used by the skill
└── README.md               # Installation and usage notes

Prerequisites

• Python 3.8+
• requests installed: pip install requests
• A QAPI account with:
  • QAPI_API_TOKEN
  • QAPI_GATEWAY_TOKEN
  • QAPI_USER_EMAIL

Environment Variables

Set these before invoking the skill:

export QAPI_API_TOKEN="your-personal-api-token"
export QAPI_GATEWAY_TOKEN="your-gateway-token"
export QAPI_USER_EMAIL="you@example.com"
export QAPI_APP_URL="qapi.qyrus.com"

The bundled client also sends scope: AI_SDK on every QAPI API request automatically.

Supported QAPI_APP_URL values:

• qapi.qyrus.com
• stg-api.qyrus.com

Any other hostname is rejected on purpose so the client does not accidentally write to the wrong environment.

Quick Start

1. Clone this repository somewhere stable.
2. Export the QAPI_* environment variables.
3. Install or reference the skill in your assistant of choice.
4. Invoke the skill and point it at the codebase you want scanned.

Installation

One-Command Install with npx skills

npx skills is the simplest install path.

Interactive install:

npx skills add QQyrus/qapi-skill

Full GitHub URL also works:

npx skills add https://github.com/QQyrus/qapi-skill

Useful non-interactive examples:

# Install globally for Claude Code
npx skills add QQyrus/qapi-skill -g -a claude-code -y

# Install globally for Codex
npx skills add QQyrus/qapi-skill -g -a codex -y

# Install globally for Cursor
npx skills add QQyrus/qapi-skill -g -a cursor -y

# Install globally for Antigravity
npx skills add QQyrus/qapi-skill -g -a antigravity -y

Why this works for this repository:

• The skills CLI supports GitHub shorthand and full GitHub URLs.
• It can prompt for the target agent if none is specified.
• It discovers skills in the repository root when the root contains SKILL.md, which this repo does.

Claude Code

Claude Code supports skill folders directly. If you do not want to use npx skills, manual install still works:

Global install:

mkdir -p ~/.claude/skills
ln -s /absolute/path/to/qapi-skill ~/.claude/skills/qapi-create-tests

Project-local install:

mkdir -p .claude/skills
ln -s /absolute/path/to/qapi-skill .claude/skills/qapi-create-tests

Usage:

/qapi-create-tests

Codex

Codex uses skill folders under $CODEX_HOME/skills and can also discover skills referenced from AGENTS.md. If you do not want to use npx skills, manual install still works:

Personal install:

mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
ln -s /absolute/path/to/qapi-skill "${CODEX_HOME:-$HOME/.codex}/skills/qapi-create-tests"

Repo-local usage through AGENTS.md:

[$qapi-create-tests](/absolute/path/to/qapi-skill/SKILL.md)

When using the skill explicitly, reference it by name or by the absolute SKILL.md path.

Cursor

Cursor does not use Codex-style skill folders. Use a project rule that tells Cursor to follow this skill, and keep this repository in the workspace so scripts/qapi_client.py remains executable. npx skills can also install directly to Cursor.

Create .cursor/rules/qapi-create-tests.mdc:

---
description: Generate and save QAPI API tests with the local qapi-create-tests skill
globs:
alwaysApply: false
---

Read `/absolute/path/to/qapi-skill/SKILL.md` and follow its execution workflow when asked to create
or save QAPI tests.

Then trigger it from Cursor Chat by naming the rule or asking Cursor to use the QAPI test creation workflow.

Antigravity

If your Antigravity setup supports reusable workspace rules or prompts, add one that points at this skill and keep this repository checked out locally. npx skills can also install directly to Antigravity.

Read /absolute/path/to/qapi-skill/SKILL.md and follow it for QAPI test generation and save flows.
Use the local scripts/qapi_client.py helper for all QAPI API calls.

The important requirement is that Antigravity can read SKILL.md and execute the bundled Python script from the same checkout.

Windsurf

Add a workspace rule such as .windsurf/rules/qapi-create-tests.md with a short instruction to read /absolute/path/to/qapi-skill/SKILL.md, then ask Windsurf to use that workflow.

GitHub Copilot / Other Prompt-Driven Assistants

These clients generally do not have a native skill install path. Reuse the skill by pasting a short instruction that points at SKILL.md, for example:

Use the qapi-create-tests skill at /absolute/path/to/qapi-skill/SKILL.md and execute it against this repository.