A collection of advanced agentic tools, skills, and workflows designed for the Antigravity AI coding assistant. This repository provides the foundation for building, auditing, and maintaining complex codebases with high precision and reliability.
These tools are designed to extend the capabilities of an AI agent, moving beyond simple code generation to structured repository analysis, systematic debugging, and professional-grade implementation planning.
Specialized modular capabilities that define hard rules and protocols for the agent.
- evidence-and-inference: A framework for evidence-based decision making and logical reasoning. Ensures the agent never "hallucinates" state but relies on file-based evidence.
- shell-and-output: A robust protocol for Windows PowerShell 7.5.4 interaction. Features mandatory output redirection, exit-code verification, and failure-pattern scanning to eliminate "silent success."
- tickets-and-logbook: A system for managing granular tasks and maintaining a durable log of progress.
High-level, multi-step procedures (slash commands) for complex objectives.
- gap-analysis: Performs a full-system audit across frontend, backend, and persistence layers to identify missing features or broken wiring.
- implementation-plan: Generates detailed, phased, and ticket-aware execution strategies.
- git-commit: Facilitates intentional, scoped, and well-documented commits.
- archeological-pass: Investigates legacy or unfamiliar codebases to reconstruct intent and define a SOTA roadmap.
To overcome the "blind spot" in standard CLI tool outputs, all agents interacting with this repo MUST:
- Redirect command output to a file (e.g.,
command > temp/output.txt). - Read the file content to verify the result.
- Clean up the temporary file immediately.
No claim about the state of a repository should be made without a corresponding view_file or grep_search call to provide proof.
- OS: Windows
- Shell: PowerShell 7+ (Core)
- Logic: Agentic workflows defined in Markdown/YAML.
Built for Advanced Agentic Coding.