Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/davidbuenov/dbv-specs-ops/llms.txt

Use this file to discover all available pages before exploring further.

dbv-specs-ops requires no package manager, no build tool, and no runtime. The entire system is a set of Markdown files your AI assistant reads automatically at session start. Getting from zero to your first disciplined /spec session takes under five minutes — follow the steps below.
Before you begin, open docs/MASTER_PROMPT.md and read it once. It is the brain of the system — the file that tells your AI assistant every rule it must follow. Understanding its structure will make every subsequent interaction more predictable.
1

Clone or download the repository

You should not clone this repository and build inside it. Instead, use it as a template for your own independent project.Option A — GitHub Template (recommended):Click the “Use this template” button on the dbv-specs-ops GitHub page. GitHub creates a brand-new repository under your account with all the files and none of the original history. The project is 100% yours from the first commit.Option B — Clone directly (for evaluation):
git clone https://github.com/davidbuenov/dbv-specs-ops.git my-project
cd my-project
rm -rf .git
git init
Option C — Download ZIP (simplest, no git required):Click the green “Code” button on GitHub → “Download ZIP”. Extract the contents and copy them to the root of your project folder.
2

Copy the framework files into your project

Whether you used the template button or the ZIP, verify these files are present at your project root after setup. The AI assistant will not find them automatically if they are nested in a subdirectory.
my-project/
├── project.config.md          ← Project identity & framework version
├── CLAUDE.md                  ← Auto-loaded by Claude Code & Cursor
├── GEMINI.md                  ← Auto-loaded by Gemini CLI & Antigravity
├── .windsurfrules             ← Auto-loaded by Windsurf
├── CHANGELOG.md               ← Version history (AI-maintained)
├── memory.md                  ← Context & decisions (AI-maintained)
├── task.md                    ← Task logbook (AI-maintained)
└── docs/
    ├── MASTER_PROMPT.md       ← The workflow brain (required)
    ├── SPECIFICATIONS.md      ← Requirements (AI-generated during /spec)
    ├── ARCHITECTURE.md        ← Tech decisions (AI-generated during /plan)
    ├── DESIGN.md              ← Visual system (optional, UI projects only)
    └── ADOPTION_PROMPT.md     ← Use this for existing projects instead
If you are integrating dbv-specs-ops into an existing repository and want to avoid overwriting files like README.md or CHANGELOG.md, copy the entire dbv-specs-ops folder as a subfolder inside your project and point your activation file there. See the Installation page for full details on both methods.
3

Pick your platform activation file

Each AI assistant platform loads context from a different file. Confirm the right one is present in your project root before opening your assistant:
PlatformFile to useHow it loads
Claude Code (CLI / VS Code / Desktop)CLAUDE.mdAutomatic at session start
CursorCLAUDE.mdAutomatic
Gemini CLIGEMINI.mdAutomatic
Antigravity (VS Code · Google DeepMind)GEMINI.mdAutomatic
Windsurf.windsurfrulesAutomatic
GitHub Copilot.github/copilot-instructions.mdAutomatic in workspace
ChatGPT / Gemini Webdocs/MASTER_PROMPT.mdManual — paste in first message
All auto-loading platforms read their activation file silently before your first message. The activation file is a lightweight pointer that tells the AI to read docs/MASTER_PROMPT.md and follow its workflow strictly.
For ChatGPT and Gemini Web, context does not load automatically. You must paste the full contents of docs/MASTER_PROMPT.md as your opening message, then follow with your task description. Skipping this step means the AI will not follow the SDD lifecycle.
4

Fill in project.config.md

project.config.md is the identity file for your project. The AI reads it at the start of every session to populate file headers, pick the right license, and confirm the framework version. Open it and replace the placeholder values with your own:
## Project Identity

- **Name:** My Awesome App
- **Description:** A cross-platform reminder system for busy teams.
- **Author / Company:** Jane Smith · https://janesmith.dev
- **License:** MIT
- **Languages:** TypeScript, CSS, HTML
- **Agent Readiness (Web):** Yes
- **Framework Version:** 2.1.0
Key fields explained:
  • Name — Used in file headers stamped on every source file the AI creates.
  • Languages — Tells the AI which header comment syntax to use (//, #, <!-- -->).
  • Agent Readiness (Web) — Set to Yes for web or API projects to have the AI scaffold robots.txt, llms.txt, .well-known/agent.json, and other AI-discovery files automatically during /build.
  • Framework Version — Identifies which version of dbv-specs-ops is installed. Used by the upgrade prompt to calculate what has changed.
You don’t have to fill this manually. If the AI detects placeholders (like [Project Name]) when it first reads project.config.md, it will run a Bootstrap Interview — proposing all six key values at once with marked assumptions for you to confirm or correct in a single reply.
5

Open your AI assistant and run /spec

With your files in place and project.config.md configured, open your AI assistant in the project directory and type:
/spec
That single command kicks off the Engineering Interview. The AI will:
  1. Read project.config.md, memory.md, and task.md silently.
  2. Analyze whether docs/SPECIFICATIONS.md already has content or just placeholders.
  3. Ask you what you want to build, who it is for, and what the key requirements are.
  4. Generate a complete draft of docs/SPECIFICATIONS.md based on your answers — no manual editing required.
For ChatGPT / Gemini Web, paste docs/MASTER_PROMPT.md first, then add:
Review task.md and start the Engineering Interview for a new project.
6

Follow the AI through the 6-phase lifecycle

After /spec produces an approved SPECIFICATIONS.md, proceed through the remaining phases in order:
/plan       ← Architect Review + atomic task breakdown + approval gate
/build      ← Incremental implementation, one slice at a time
/test       ← Unit/integration tests; task not done without passing tests
/code-simplify  ← Refactor for clarity + mandatory Security Review
/ship       ← Versioning, CHANGELOG, README update, git tag proposal
You can check progress at any point by reading task.md — the AI keeps it up to date with a checklist of completed and pending steps, plus a Context Snapshot so any session can resume exactly where the last one ended.After /ship, the AI updates memory.md with any new Architecture Decision Records or lessons learned, ensuring the next feature cycle starts with full context.

​
What happens next

Once /ship completes, your project has a versioned release, a populated CHANGELOG.md, an updated README.md, and a walkthrough.md summarizing everything built and tested. The cycle then loops — the next feature starts with /spec, and memory.md carries all prior context forward, eliminating AI amnesia across sessions.

Installation

Learn about fresh-project vs. existing-project setup methods.

Introduction

Understand the Harness equation and the full framework architecture.

Build docs developers (and LLMs) love

Get started for freeTalk to us