Using dbv-specs-ops with Gemini CLI and Antigravity

A single activation file — GEMINI.md — serves both Gemini CLI and Antigravity (VS Code · Google DeepMind). Both tools read GEMINI.md automatically from the workspace root at the start of every session, giving them the full Spec-Driven Development context before any code or plan is proposed. Antigravity extends this base with three native features: Planning Mode with versioned artifacts, Knowledge Items for context recovery, and Context Snapshots for cross-session persistence.

How GEMINI.md Is Auto-Loaded

Both Gemini CLI (gemini terminal command) and Antigravity detect GEMINI.md at the project root and read it as part of their session initialization. No manual pasting is required on either platform. The file is a strong pointer: rather than duplicating workflow logic, it directs the AI to docs/MASTER_PROMPT.md as the single authoritative source for the full SDD lifecycle.

The ANTIGRAVITY.md file documents the Antigravity-specific extended features. It is read as reference documentation — GEMINI.md remains the primary auto-loaded activation file for both platforms.

Files Read at Session Start

GEMINI.md instructs the AI to read these seven files before proposing any code or plan:

File	Purpose
`project.config.md`	Project identity: name, author, license and file header template
`docs/MASTER_PROMPT.md`	Mandatory workflow, rules and boundaries
`docs/SPECIFICATIONS.md`	Current project requirements
`docs/ARCHITECTURE.md`	Stack and technical decisions
`docs/DESIGN.md`	Visual design system: color tokens, typography, components and philosophy (if it exists)
`memory.md`	Context and Decisions: Qualitative knowledge (ADRs, lessons learned, active context)
`task.md`	Current state + Context Snapshot

Core Rules

The core directive in GEMINI.md is identical to the one in CLAUDE.md, ensuring consistent behavior across platforms:

Read docs/MASTER_PROMPT.md and follow its workflow strictly. If you detect contradictions between the prompt and project specs, halt and report before proceeding.

All initialization (Bootstrap), state checking (Specs Check), lifecycle (Workflow) and coding standards are defined centrally in MASTER_PROMPT.md to avoid cognitive redundancy.

Platform-Specific Behaviors

Gemini CLI
Antigravity

Gemini CLI reads GEMINI.md automatically when you run the gemini command inside a project directory that contains the file. The base SDD workflow defined in docs/MASTER_PROMPT.md applies in full. No additional configuration is needed.Setup steps:

Place GEMINI.md at the project root

Copy GEMINI.md from the dbv-specs-ops template to the root of your project. Gemini CLI will find it automatically on the next session start.

Fill in project.config.md

Complete the project identity fields. If left with placeholders, Gemini will run the Bootstrap interview on first session start.

Start a Gemini CLI session

Run gemini in your terminal from the project root. The full SDD context is loaded automatically.

Type /spec to begin

Send /spec to start the Engineering Interview. Gemini will analyze the project and propose a specification draft before writing any code.

Antigravity (VS Code · Google DeepMind) reads GEMINI.md using the same auto-load mechanism as Gemini CLI, but adds three native capabilities on top of the base SDD workflow.Antigravity-Specific Behaviors:

Planning Mode

When creating a plan, Antigravity’s native Planning Mode should be activated. The planning artifacts — implementation_plan.md, task.md, and walkthrough.md — must be created inside the project workspace root, not only in the conversation brain directory. This ensures they are versioned with the project and accessible to the team and to other AI platforms that read from the filesystem.

Knowledge Items (KIs)

After completing a significant milestone, Antigravity can create a Knowledge Item (KI) that summarizes the full project context. KIs enable seamless context recovery in future sessions without needing to reread all documentation files from scratch. When a milestone is reached, the AI will offer to create a KI.

Context Snapshot

At the end of each session, or before reaching a conversation context limit, Antigravity writes a Context Snapshot to task.md. The snapshot records the exact next step needed to resume work instantly in a future session — eliminating the ramp-up cost of re-establishing context from zero.Setup steps:

Place GEMINI.md at the project root

Copy GEMINI.md from the dbv-specs-ops template to your project root. Antigravity will auto-load it at session start.

Review ANTIGRAVITY.md for reference

The ANTIGRAVITY.md file documents the extended behaviors in detail. You do not need to copy it anywhere — it serves as documentation about how to take full advantage of Planning Mode, Knowledge Items, and Context Snapshots.

Open the project in VS Code with Antigravity

Open your project folder in VS Code. Antigravity reads GEMINI.md from the workspace root at session start.

Activate Planning Mode for complex tasks

For any task that qualifies as complex (affects more than three files, involves authentication or sensitive data, or is estimated at more than 150 new lines), ask Antigravity to activate Planning Mode. It will create the plan artifacts in the workspace root so they are versioned and visible to the team.

Request a Knowledge Item after milestones

After completing a significant feature or phase, ask Antigravity to create a Knowledge Item. In future sessions, loading the KI replaces manually re-reading all seven startup files.

If you maintain a team working across multiple AI platforms — for example, one developer using Antigravity and another using Claude Code — the shared project files (task.md, memory.md, docs/SPECIFICATIONS.md) keep everyone in sync automatically. Each platform reads the same source of truth.

Get Started

Core Concepts

The SDD Workflow

Platform Setup

Advanced

Using dbv-specs-ops with Gemini CLI and Antigravity

How GEMINI.md Is Auto-Loaded

Files Read at Session Start

Core Rules

Platform-Specific Behaviors

Planning Mode

Knowledge Items (KIs)

Context Snapshot

Build docs developers (and LLMs) love

Get Started

Core Concepts

The SDD Workflow

Platform Setup

Advanced

Documentation Index

​How GEMINI.md Is Auto-Loaded

​Files Read at Session Start

​Core Rules

​Platform-Specific Behaviors

​Planning Mode

​Knowledge Items (KIs)

​Context Snapshot

Build docs developers (and LLMs) love

How GEMINI.md Is Auto-Loaded

Files Read at Session Start

Core Rules

Platform-Specific Behaviors

Planning Mode

Knowledge Items (KIs)

Context Snapshot