Connected-Mate

Specdrive — Testing skill for Claude Code

Testing community

Spec-driven development, enforced.

How to install Specdrive

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Connected-Mate/specdrive and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Specdrive does

Spec-driven development, enforced. An MCP server + Mac app that forces AI coding agents through specs, scenarios, proof-gated tasks and independent review — because prompts get forgotten, tools can't be.

Alternatives in Testing

Notion Spec To Implementation — Convert Notion specs into linked implementation plans and tasks 14.6k ★
Test Scenarios — Create comprehensive test scenarios from user stories 7.8k ★
Regen API — Regenerate the OpenAPI spec and web API client after server API changes 2.5k ★

README

SpecDrive

**Your idea, built properly.** SpecDrive is a free, open-source Mac app that makes spec-driven development effortless for people who don't code.

**Download the app**: [specdrive website](https://connected-mate.github.io/specdrive-site/) · MIT licensed · built by [Connected Mate](https://github.com/Connected-Mate)

Why an MCP and not a "skill"/prompt file? Prompts live in the agent's context and fade as it fills up — that's exactly when agents drift. SpecDrive puts the discipline in the *tools*: the to-do list is enforced, out-of-phase moves fail, "done" requires proof, and the board (not the context) carries the memory. Works with any MCP-capable agent, for any team.

You describe what you want to build — in your own words. Your AI coding agent (Claude Code, Cursor, Windsurf, Gemini CLI, Codex CLI…) fills a beautiful live board with your project's specs, challenges them, researches them online, flags the genuinely hard parts, sketches the screens, then builds the product step by step — and you watch every card appear in real time.

How it works

Connect — SpecDrive detects the AI agents installed on your Mac and registers its local MCP server with them in one click.
Talk — Copy the starter prompt into your agent and describe your idea. Every answer becomes a spec card on your board, live.
Loop — The board walks you through the quality loop, handing you the exact prompt for each step (best pasted into a fresh agent session):
- Idea → capture everything you know
- Challenge → a fresh session hunts for holes and contradictions
- Research → web research: similar products, reusable building blocks, pitfalls
- Hard parts → a pre-mortem ranks difficulty and plans fallbacks; hard topics get dedicated deep-dive sessions
- Plan → architecture, screen sketches (wireframes) and a small-steps build plan
- Build → strict discipline: one task at a time, verified, checked off on your board