Skip to content
This repository was archived by the owner on Jul 9, 2026. It is now read-only.

GETTING_STARTED.md

Latest commit

 

History

History
139 lines (95 loc) · 6.29 KB

GETTING_STARTED.md

File metadata and controls

139 lines (95 loc) · 6.29 KB
Raw

TETHER Antigravity User Manual

The definitive guide to disciplined AI-human collaboration.

TETHER (v3.0) is a safety cage and workflow protocol for Google Antigravity. It transforms the AI from a simple autocomplete engine into a disciplined engineering partner that follows a strict "Plan-Then-Execute" mandate.

🏛️ Core Philosophy: The Safety Cage

TETHER isn't just a set of templates; it's a guardrail system designed to prevent "AI Drift"—the tendency for LLMs to cut corners, hallucinate, or lose track of project goals.

Mandatory Guardrails

  • Plan-Then-Execute: The AI must create an implementation plan and wait for your approval before writing a single line of code.
  • Metronome Anti-Shortcut Checks: Every 5 tool calls, the AI pauses to self-audit. It checks for signs of "cutting corners" (like bulk file deletes or using placeholders).
  • 60% Context Budget: To keep the AI sharp, TETHER enforces a context limit. When the AI reaches 60% of its memory capacity, it must archive the current phase and clear its head.
  • No Placeholders: You will never see // ... rest of code. TETHER AI writes complete, functional code or breaks the task into smaller fragments.

🚀 Getting Started

1. Installation

To bring TETHER into your project:

# In your project root
git clone https://github.com/Rauglothgor/tether-antigravity.git .tether-temp
Copy-Item -Recurse .tether-temp/* . -Force
Remove-Item -Recurse .tether-temp

2. Initialization

Once the files are in place, tell the AI:

"Initialize TETHER for this project."

This will create your core planning directories:

  • .planning/: Where your goals and roadmaps live.
  • .tether/: System configuration and internal memory.
  • 00_ARCHITECT_BRAIN.md: The constitution the AI follows.

🛠️ The TETHER Workflow

A standard TETHER development cycle follows four distinct phases:

  1. Specify (/specify): Define exactly what you want. This creates a FEATURE_SPEC.md.
  2. Plan: The AI creates an implementation_plan.md. You review and approve.
  3. Execute: The AI writes the code, following strict quality standards and auto-linting.
  4. Verify: The AI runs tests and creates a walkthrough.md to prove it works.
  5. Archive (/archive-phase): Lessons are extracted, and context is cleared for the next task.

⌨️ Slash Command Reference

Commands that give you total control over the engine.

Command Action Best Used When...
/status Shows project health & Traffic Lights. You want to see if the AI is "tired" or off-track.
/help Displays the command reference. You forget what a command does.
/specify [feature] Generates a detailed requirement spec. You're starting a new feature.
/debug [issue] Starts a structured investigation. Something broke and you don't know why.
/investigate [q] Spawns a sub-agent for research. You need to research a library without making edits.
/decision [topic] Logs an Architectural Decision (ADR). You're choosing between two different approaches.
/godmode [goal] Enables full autonomous building. You have a clear goal and want the AI to handle it end-to-end.
/archive-phase Compresses context and saves lessons. You've finished a major task and the AI feels "slow".
/evergreen-verify Scans for deprecated documentation. You want to ensure your docs match the current code.
/rollback Reverts the last set of changes. The AI went in the wrong direction.

📂 Key Artifacts

Understanding the files that power your workspace.

The Planning Layer (.planning/)

  • PROJECT.md: The "North Star". Defines the tech stack and high-level architecture.
  • ROADMAP.md: The chronological path. Keeps work organized in phases.
  • DECISIONS.md: The "Why". Tracks every major technical choice.
  • FEATURE_SPEC.md: The "What". Detailed requirements for a specific feature.

The Logic Layer (.tether/)

  • STATE.json: The AI's internal metrics (accuracy, memory usage, health).
  • FAILURE_MODES.md: A registry of known risks and how to avoid them.
  • SKILLS_INDEX.md: Reusable patterns the AI can "learn" from.
  • AUDIT.md: A log of every approval and major action.

🧠 Global Memory (.thoughts)

TETHER allows you to share learnings across different projects.

By creating a symbolic link named .thoughts that points to a central directory on your PC, Project A's mistakes can become Project B's wisdom. See GLOBAL_MEMORY_SETUP.md for setup instructions.

⚡ Advanced Power Features

God Mode

Run /godmode [goal] to trigger the BMad Method (Brain → Map → Act → Done). The AI will autonomously:

  1. Specify the feature.
  2. Build the architecture.
  3. Execute the code.
  4. Verify with a browser sub-agent.
  5. Archive the result.

Swarm Architecture

For massive tasks, TETHER can spawn multiple sub-agents to work in parallel. Mention "Swarm Mode" in your plan to activate multi-agent orchestration.

🚦 Understanding Traffic Lights

The /status command uses a traffic light system to report on system health:

  • 🟢 Nominal (85-100): AI is sharp, context is clean. Proceed with complex tasks.
  • 🟡 Warning (70-84): Context is getting heavy. The AI might start missing details. Consider /archive-phase.
  • 🔴 Critical (<70): High risk of hallucinations. Stop work immediately and run /archive-phase.

❓ FAQ & Common Pitfalls

Q: The AI is ignoring my instructions. A: Check 00_ARCHITECT_BRAIN.md. If the rules are being violated, remind the AI to "Consult the Architect Brain."

Q: I lost my implementation plan. A: Plans are always stored in the artifacts directory. Run /status to find the path to your current task log.

Q: Can I skip the planning phase? A: No. TETHER is designed to prevent technical debt. Skipping the plan is a "Hard Violation" and will degrade the AI's performance.

Welcome to TETHER. Discipline is freedom.