antigravity-protocol

Enforces token-efficient coding workflows with chunked edits and plan-first execution.

Updated Jun 30, 2026
One-click install
npx skills add https://github.com/kaitoartz/dotfiles --skill antigravity-protocol-kaitoartz
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: antigravity-protocol
Source: https://github.com/kaitoartz/dotfiles/tree/main/dot_gemini/config/skills/antigravity-protocol
Command: npx skills add https://github.com/kaitoartz/dotfiles --skill antigravity-protocol-kaitoartz

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI coding assistants often waste tokens on verbose preambles, full-file rewrites, and exploratory guessing. This Skill constrains the assistant to a strict, low-token workflow with targeted edits and structured planning for software development tasks. ## Core Features & Use Cases - Three Execution Modes: Investigatory mode for code questions, Fast Path for small fixes, and Strict Planning Mode for large features requiring an approved implementation plan. - Chunk-Based Editing: Mandates search-and-replace edits against exact line numbers instead of outputting full file contents. - Persistent Knowledge: Creates system_architecture.md or .cursorrules save-state files so future sessions reuse architectural context. - Use Case: When asked to implement authentication across a codebase, the assistant silently traces dependencies, writes an implementation_plan.md, waits for approval, then executes via a task.md checklist and verifies with a build command. ## Quick Start Ask the assistant to implement a new feature or fix a bug and it will follow the antigravity protocol with a plan-first, low-token workflow.

Frequently Asked Questions about antigravity-protocol

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce token usage when an AI writes code?▼

Use a protocol that bans full-file rewrites and conversational preambles, requiring chunk-based search-and-replace edits against exact line numbers. Large tasks go through a plan artifact first, and substantial output is moved into local Markdown files instead of chat.

How to make an AI assistant plan before coding large features?▼

Strict Planning Mode requires the assistant to silently trace dependencies, write an implementation_plan.md listing files as [NEW], [MODIFY], or [DELETE], then halt for user approval before executing any code changes via a task.md checklist.

Can this workflow work with Flutter or npm projects?▼

Yes, the verification phase runs the project's own build or test command such as npm run build or flutter test after edits. Terminal commands are restricted to tests, servers, and package installation only.

Why does the assistant avoid terminal commands like cat and grep?▼

Generic shell commands for reading or editing files consume excessive tokens and produce imprecise results. The protocol enforces a tool hierarchy prioritizing built-in read, search, and edit tools, reserving the terminal for builds and tests.

When should I not use a strict low-token coding protocol?▼

Avoid it for open-ended exploration, learning-oriented explanations, or tasks needing detailed discussion of tradeoffs. The protocol suppresses explanatory output and stops to ask questions rather than reasoning through ambiguity at length.